CN101770579B - A Feature Description Method for Image Objects - Google Patents

Abstract

The invention discloses a feature description method of an image target, which belongs to the basic technology in the field of pattern recognition. The method is as follows: divide the image target into multiple geometric figures; obtain the feature information of each geometric figure; further construct the feature information describing the image target according to the obtained set of geometric figure feature information, and realize the accurate identification of the main features of the image target describe. The method for acquiring image target feature information is simple, and the amount of calculation is greatly reduced. In the case of rotation, translation, and scaling, feature information that accurately describes the image target can be obtained.

Description

A Feature Description Method for Image Objects

technical field

The invention relates to the field of pattern recognition, in particular to a feature description method and device for an image object. the

Background technique

Image target recognition is the key technology of graphic and image recognition in the field of pattern recognition. Image target recognition technology plays an important role in artificial intelligence, computer vision, robotics, image target recognition, optical character reader (Optical Character Reader, OCR), military and other high-tech fields. the

An image object refers to an image with clear and identifiable boundaries, composed of multiple geometric figures, or a partial image of an image. the

Image object recognition is realized based on the feature description of image objects. At present, there are mainly two methods of feature description for image objects:

The first one, based on the Fourier descriptor theory of integral transform, transforms the image information from the spatial domain to the frequency domain, and uses the obtained frequency domain feature vector set to realize the overall description of the image target;

The second is based on the invariant moment theory, the Riemann double integral of the density function on the image, the moments of different orders have different physical meanings, and the overall description of the image target is realized by using the moment parameter set. the

In the above two methods, the feature vector set and the moment parameter set in the frequency domain are invariant to translation, scaling and rotation. When the density function on the image is equal to 1, the frequency domain feature vector set and the moment parameter set realize the overall description of the image target. the

Get feature information. Many application examples also illustrate the major flaws existing in the mathematical description of image objects in the prior art. The specific analysis is as follows:

Using the above two methods to describe the features of the image target, although it is universal, but due to some technical defects, the practical application of the above two methods is greatly limited, the specific analysis is as follows:

First, the description of image target features by frequency-domain feature vectors or moment parameters is a relatively rough description method with more uncertainty. There is no relationship between frequency-domain feature vectors or moment parameters and the main features of image targets. Reliable theoretical basis, and it is impossible to have a credible corresponding relationship. Therefore, the accuracy of feature description of image targets using frequency-domain feature vectors or moment parameters is low;

Second, obtaining image target feature information through frequency-domain feature vectors or moment parameters requires a large amount of calculation;

Thirdly, obtaining image object feature information through frequency-domain feature vectors or moment parameters cannot describe the common features of a class of image objects. the

Contents of the invention

In view of the problems existing in the prior art, the object of the present invention is to provide a feature description method of an image object, which is used to solve the defects that the prior art cannot describe the image object accurately and effectively, and the calculation of feature information requires a large amount of calculation. the

In order to achieve the above object, the technical solution proposed by the present invention is: a feature description method of an image target, the method comprising:

a. Search for pixels similar to the boundary optical parameters in the image target, form multiple closed curves according to the searched pixels, and use the figure formed by each closed curve as a geometric figure;

b. Calculate the boundary curve of the obtained geometric figure according to a certain direction, and obtain the curvature information of each point on the boundary curve; obtain the reference point of the boundary curve of the geometric figure; obtain the boundary curve of the geometric figure, Mark the feature point of the change feature; use the reference point as the pole of the polar coordinate, calculate the extreme value and polar angle of the polar coordinate vector of the feature point; calculate and obtain the feature point type code, additional feature code and curvature radius according to the curvature information; The extremum, polar angle, feature point type code, additional feature code and radius of curvature of the polar coordinate vector constitute feature point information, and form feature information describing the shape of the geometric figure according to the calculated feature point information;

c. Determine the feature information of the image object according to the obtained set of geometric figure feature information, and describe the feature of the image object. the

An embodiment of the present invention provides a feature description device for an image object, the device comprising:

The graphics division unit is used to search for pixels close to the boundary optical parameters in the image target, form a plurality of closed curves according to the searched pixels, and use the graphics formed by each closed curve as a geometric figure;

Graphic feature determining unit, including curvature obtaining unit, reference point obtaining unit, feature point obtaining unit and feature information obtaining unit; said curvature obtaining unit is used to obtain the boundary curve of the geometric figure according to a certain direction Carry out the calculation, and obtain the curvature information of each point on the boundary curve; the reference point obtaining unit is used to obtain the reference point of the boundary curve of the geometric figure; the feature point obtaining unit is used to obtain the On the boundary curve of the geometric figure, mark the feature point of the change feature; the feature information obtaining unit is used to use the reference point as the pole of the polar coordinate to calculate the extreme value and polar angle of the polar coordinate vector of the feature point; according to The curvature information is calculated to obtain the feature point type code, additional feature code and radius of curvature; the feature point information is composed of the extreme value, polar angle, feature point type code, additional feature code and curvature radius of the polar coordinate vector, and the calculated feature point The information constitutes characteristic information describing the shape of the geometric figure;

An image feature determining unit, configured to determine feature information of the image object according to a set of geometric figure feature information acquired by the graphic feature determining unit, and describe the feature of the image object. the

In the present invention, by dividing the image target into a plurality of geometric figures, obtaining the feature information of each geometric figure, and determining the feature information of the image target according to a group of geometric figure feature information obtained, the process of obtaining the image target feature information is only Computing and processing a small amount of data in the image effectively reduces the amount of calculation compared with the existing technology, and because the features of the multiple geometric figures contained in the image target can accurately reflect the characteristics of the image target, so using the image target A set of geometric figure feature information is included to describe the features of the image target, which can effectively improve the accuracy of describing the feature of the image target. the

Description of drawings

Fig. 1 is the schematic flow chart of the method that the embodiment of the present invention provides;

Fig. 2 is a schematic flow chart of region division in the embodiment of the present invention;

Fig. 3 is a schematic flow chart of determining regional feature information in an embodiment of the present invention;

Fig. 4 is the image object in the example of the present invention;

Fig. 5 is the schematic diagram of the geometry figure that image object comprises in Fig. 4;

Figure 6A is a schematic diagram of the feature information of the geometric figure t1 in Figure 5;

Figure 6B is a schematic diagram of the feature information of the geometric figure t2 in Figure 5;

Figure 6C is a schematic diagram of the feature information of the geometric figure t3 in Figure 5;

Figure 6D is a schematic diagram of the feature information of the geometric figure t4 in Figure 5;

Figure 6E is a schematic diagram of the feature information of the geometric figure t5 in Figure 5;

Figure 6F is a schematic diagram of the feature information of the geometric figure t6 in Figure 5;

Figure 7A is a schematic diagram of the feature information of the first area in Figure 5;

Figure 7B is a schematic diagram of the characteristic information of the second area in Figure 5;

Figure 7C is a schematic diagram of the characteristic information of the third area in Figure 5;

Fig. 8 is a schematic structural diagram of a device provided by an embodiment of the present invention. the

Detailed ways

In our daily life, we are recognizing image objects anytime and anywhere, and we are used to it and don't take it seriously. However, in the field of pattern recognition, the ability of machines to recognize image objects is still very low. the

Both humans and animals with visual ability have the ability to recognize image targets, and they have this ability consciously or unconsciously in the basic activities of survival. This ability comes from imagery thinking. Imagery thinking is through perceiving imagery information, using imagery knowledge (imagery, imagery, experience, etc.) Compared with the existing image target recognition technology, this thinking activity does not use complicated mathematical theories, nor does it have a lot of complicated calculations, but it is simple, fast and effective. Having the IQ of an animal can correctly identify the image target, which objectively shows that there is a simplified and efficient image target description and recognition method. the

It is the core idea of the present invention to simulate image thinking and design a feature description method of image objects. According to human's direct perception of image target recognition, it can be seen that image thinking realizes the memory (description) of image targets by directly perceiving the boundary contour features of multiple geometric figures that constitute image targets. The characteristic points of the boundaries of each geometric figure are calculated by mathematical methods, and the information of the characteristic points is obtained. Learn the advantages of image thinking and overcome the shortcomings that cannot be quantified and described. The feature information of multiple geometric figures is used to accurately describe the main features of the image target. the

Another main idea of the present invention is to simulate image thinking and extract common features of a class of image objects. Image thinking is particularly profound in the memory (description) of the layers, positional relationships, and convex and concave features of the geometric figures that make up the image. By simulating this feature, the statistical information describing the overall characteristics of the image is extracted from the image feature information to realize the classification and retrieval of images. the

Referring to Fig. 1, the feature description method of the image target provided by the embodiment of the present invention comprises the following steps:

Step 10: Divide the image target into multiple geometries;

Step 11: For the divided geometric figures, obtain the feature information of each geometric figure;

Step 12: Determine the feature information of the image object according to the acquired feature information of a group of geometric figures, and describe the feature of the image object. the

The following is a detailed description of each of the above steps:

In step 10, the image target is divided into multiple geometric figures, and its specific implementation method is:

Pixels with similar boundary optical parameters are searched in the image target, and the searched pixels form multiple closed curves, and the figure formed by each closed curve is regarded as a geometric figure. the

The feature information of the geometric figure in step 11 includes feature point information of the geometric figure, reference point information of the boundary curve of the geometric figure, and basic retrieval information of the geometric figure. For the specific determination method of the reference point information, the feature point information and the basic search information, please refer to the Chinese patent with the application number "200710303995.9". For the convenience of understanding, the terms involved are explained as follows: in the previous patent "Plane geometric shape" is an abstraction of the "geometric figure" in this patent. The shape formed by the boundary of the "geometric figure" in this patent is the "plane geometric shape". The optical parameters of the "geometric figure" describe the optical characteristics of the geometric figure. In this patent, "geometric figures" are used uniformly without distinction. in:

The method for obtaining the feature point information of the geometric figures disclosed in the patent is as follows:

First, calculate the boundary curve of the geometric figure according to a certain direction, and obtain the curvature information of each point on the boundary curve;

Then, obtain the reference point of the boundary curve of the geometric figure;

Then, obtain the feature points on the boundary curve of the geometric figure that mark the changing features;

Finally, feature point information is calculated according to the feature point, the reference point, and the curvature information. the

The feature point information of the calculated geometric figure includes the following content: the polar coordinate vector (including extreme value and polar angle) of the feature point on the boundary curve of the geometric figure with the reference point of the geometric figure as the pole (including extreme value and polar angle), the The radius of curvature, the type code of the feature point, and the additional feature code of the feature point. . the

The feature point information is denoted as tz(l, s, t, m, r), wherein, l, s are extremum and polar angle in the polar coordinate vector information, r is the radius of curvature of the feature point, m is the additional feature code of the feature point, and t is the type code of the feature point. the

The method for determining the reference point of the boundary curve of the geometric figures disclosed in the patent is as follows:

The reference point is obtained by calculating the boundary or area information of the geometric figure, and the reference point has invariance relative to the boundary in the case of rotation, translation and scaling. The specific calculation method is not limited. Denote the reference point of the boundary curve of the geometric figure as pc;

The determination method of the basic retrieval information of the geometric figure disclosed in the said patent is as follows:

Extracting various statistical information from the set of feature point information constitutes the basic retrieval information describing the overall characteristics of the geometric figure, and the basic retrieval information includes the following contents:

The maximum vector of feature point information (the serial number of feature point information with the maximum vector is denoted as o _t1 );

The ratio of the maximum vector of feature point information to the minimum vector extremum (denoted as k1);

The number of various types of feature points. The types of feature points include convex points, concave points, and tangent points. Including: the number of convex points on the boundary curve (denoted as n _t ), the number of concave points on the boundary curve (denoted as n _z ), the number of tangent points on the boundary curve (denoted as n _q );

The total number of feature points (denoted as n _s );

The accumulated sum of the absolute value of the rotation angle of the feature point information vector (denoted as jd1);

The total length of the boundary (denoted as zc);

The ratio of the cumulative length of various types of boundary curve segments to the total length of the boundary curve. The types of boundary curve segments include straight lines, circular arcs, monotonously increasing arcs of curvature, monotonously decreasing arcs of curvature, etc. The ratio of the cumulative length of straight lines to the total length of boundary curves (record zx), the ratio of the cumulative length of the arc to the total length of the boundary curve (denoted as yh), the ratio of the cumulative length of the monotonously increasing curvature arc to the total length of the boundary curve (denoted as dz), the cumulative length of the monotonically decreasing curvature and the total length of the boundary curve The ratio of the total length of the boundary curve (denoted as dj). the

Optical parameters on the boundary curve of the geometric figure (denoted as bg);

Record the basic retrieval information of geometric figures as js(k ₁ , n _t , n _a , n _z , n _y , n _q , n _s , o _t1 , jd ₁ , zc, zx, yh, dz, dj, bg) .

The feature information describing the geometric figure is composed of the boundary curve reference point, a set of feature point information and basic retrieval information. the

So far, the feature information of the geometric figure can be recorded as TZ(pc, js, t _Z1 , t _Z2 , t _{Z3 .} . . t _Zn ), where n is the number of feature points of the geometric figure.

In step 12, the characteristic information of the image object is determined according to the obtained characteristic information of a group of geometric figures, and the characteristics of the image object are described. Specifically, there are two implementation methods as follows:

The first one uses the feature information of the set of geometric figures to directly describe the image target:

The feature information of each geometric figure describes the local features of the image target, a set of feature point information in the feature information of the geometric figure describes the change characteristics of the boundary of the geometric figure, and the basic retrieval information in the feature information of the geometric figure describes the overall characteristics of the geometric figure. The optical parameters in the feature information describe the optical characteristics of the geometric figures, the reference points in the feature information of the geometric figures describe the positional relationship between the geometric figures, and the feature information of a group of geometric figures describes the main features of the image target. the

This method achieves an accurate description of the main features of the image target, but the feature information of the geometric figure has no priority, no sequence, and no hierarchy, which is not conducive to understanding and use. the

The second is to divide the region where the image target is located into multiple sub-regions of different levels according to the mutual inclusion relationship of a group of geometric figures, and determine the characteristic information of each region. Describe the main features of the image target:

Regional division should determine the level and serial number of the area. The region is expressed as Q _Nm , where N represents the region level, and m represents the region number; the region division must also determine the number of independent geometric figures contained in the region; set the serial number of the independent geometric figures; determine the independent geometry that defines the next-level region in the region The serial number of the graph. The highest level of the area is set to 1, and each time the level of the area is lowered, the value of the marked level is increased by one. A geometric figure that is not surrounded by other geometric figures in the current area is called an independent geometric figure, and an independent geometric figure that surrounds other geometric figures in the current area can define the next-level area.

The region division process is shown in Figure 2, including the following steps:

Step 20: Set the area where the image object is located as the highest-level area, set the current area level N=1, there is only one highest-level area, and set the number of sub-areas S _N =1 at the current area level. Set the serial number of the independent geometric figure bounding the current area to P _N [1]=0. Set the area number m=0, set the number h=0 of the independent geometric figure; set step 21 to start dividing the first sub-area of this level k=1; set the number of sub-areas of the next level S ₀ =0.

Step 21: In the kth sub-area of level N of the current area, determine the independent geometric figures in this area, and determine the independent geometric figures in this area that define the next-level area. The subregion is bounded by the boundaries of the individual geometric figures numbered P _N [k]. When P _N [k]=0, the sub-area is the area where the image object is located. In this area, C ₀ geometric figures satisfying the conditions of independent geometric figures are obtained by boundary search or other methods, and these C ₀ geometric figures are sorted by boundary length and marked as independent with sequence numbers h+1 to h+C ₀ Geometry d _h+1 ...d _h+C0 .

Search separately according to the order of the mark serial number, and judge whether d _h+1 ... d _h+C0 independent geometric figure demarcates whether the area contains geometric figures, when the independent geometric figure with the serial number j (h+1≤j≤h+C0) contains geometric figures In graphics, execute S ₀ =S ₀ +1; p ₀ [S ₀ ]=j. After the search and judgment are completed, the area number m=m+1; C _m =C ₀ ; the current area information Q _Nm is obtained including C _m independent geometric figures d _h+1 ... d _h+Cm . Then the serial number of the independent geometric figure h=h+C _m .

Step 22: Execute k=k+1; judge if k>S _N , the sub-areas of this area level have been divided, go to step 23, otherwise, there are undivided sub-areas at this area level, go to step 21.

Step 23: Determine the number S ₀ of the independent geometric figures defining the next-level area determined in step 21. When S ₀ >0, it means that there is a sub-area of the next level to be divided. Execute, set the first sub-area k=1 of the next area level; set the marked value N=N+1 of the next area level; set the number of sub-areas S _N =S ₀ at the area level N; set the bounded area The number of independent geometric figures P _N [i]=p0[i](1≤i≤S ₀ ) of each sub-area whose level is N; the number of sub-areas whose level is N+1 is initialized S ₀ =0; go to step twenty one. Otherwise, the division process ends.

After the region is divided, N is the number of levels of the region division where the image object is located; S _j (1≤j≤N) is the number of sub-regions contained in each area level; m is the region division where the image object is located The number of areas; C _j (1≤j≤m) is the number of independent geometric figures contained in each area; h is the number of geometric figures divided by the area where the image target is located; each geometric figure is in a specific area called independent geometry. d _j (1≤j≤h) is h independent geometric figures arranged in the hierarchical relationship divided by region.

The process of obtaining regional feature information is shown in Figure 3, including the following steps:

Step 30: Set the region where the image object is located as the current region for obtaining region feature information. Set the area level N _s =1 of the current area; set the area number m _s =1 of the current area; set the area number discrimination parameter S _c =S ₁ of the current area level;

Step 31: Determine the independent geometric figures contained in the current area; according to the result of area division, the number of independent geometric figures in the current area C ₀ =C _ms ; when the area number m _s =1, the start of the independent geometric figures in the current area The sequence number g=1, when the region sequence number m _s >1, the start sequence number of the independent geometric figure in the current region is g=g+C _ms . The continuous C ₀ independent geometric figures d _g ... d _g+C0-1 starting from the serial number g are independent geometric figures contained in the current area.

Step 32: Obtain the reference point of the current area, denoted as qpc;

The reference point of an area is a point that maintains the same relative positional relationship with the reference point of each independent geometric figure in the area under the conditions of rotation, translation, and scaling. The reference point can be calculated by using one or more of the reference points, boundary information, and area information of independent geometric figures in the area, or by using area boundary information and/or area information. The method of calculating the quantity is illustrated with an example:

First, the feature information of the geometric figure corresponding to the independent geometric figure in the current area is relabeled as dTZ _b according to the serial number of the independent geometric figure. In the current area, get the total length and reference point of the boundary curve of each independent geometric figure contained;

Then, calculate the horizontal coordinate value qpc.x and vertical coordinate value qpc.y of the area reference point according to the obtained information according to the following formula:

$\mathrm{<span\; class="notranslate">\; qpc</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; g</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; dZ</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; js</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; zc</span>}<span\; class="notranslate">\; ;</span>$

$\mathrm{<span\; class="notranslate">\; qpc</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; g</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; dZ</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; pc</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; dZ</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; js</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; zc</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; QZ</span>}<span\; class="notranslate">\; ;</span>$

$\mathrm{<span\; class="notranslate">\; qpc</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; g</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; dZ</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; pc</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; dZ</span>}}_{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; js</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; zc</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; QZ</span>}<span\; class="notranslate">\; ;</span>$

Among them, dTZ _b is the feature information of the geometric figure corresponding to the independent geometric figure; the serial number b of dTZ _b in the current area is (g≤b≤n0); where n0=g+C0-1, C0 is the current area contains The number of independent geometric figures of . dTZ _b .pc.x is the horizontal coordinate value of the reference point of the independent geometric figure with the serial number b contained in the current area, and dTZ _b .pc.y is the vertical coordinate of the reference point of the independent geometric figure with the serial number b contained in the current area value, dTZ _b .js.zc is the total length of the boundary curves of independent geometric figures with serial number b contained in the current area.

Step 33: For each independent geometric figure in the current area, determine the characteristic information of the independent geometric figure, which is recorded as DTZ _b , and the value of the serial number b is (g≤b≤g+C0-1);

The determination method of DTZ _b is as follows:

Obtain the reference point dTZ _b.pc of the independent geometric figure whose sequence number is b, and take the current area reference point qpc as the polar coordinate vector of the pole, which is called the orientation vector (denoted as wl);

The obtained orientation vector wl, the characteristic information of the geometric figure corresponding to the independent geometric figure (denoted as dTZ _b ), and the characteristic information of the area defined by the independent geometric figure (denoted as QTZ) are used as the characteristic information of the independent geometric figure, Denote as DTZ _b (wl, dTZ _b , QTZ).

Among them, the orientation vector wl describes the relative position relationship of the independent geometric figure in the area; dTZ _b describes the boundary characteristics of the independent geometric figure; QTZ describes the internal characteristics of the area bounded by the independent geometric figure, that is, the characteristics of the next-level area. When no other geometry is contained within the bounding area of the independent geometry, the QTZ will not be included in DTZ _b .

Since the regional characteristic information is obtained by level and region, the region defined by the current independent geometric figure must be at the next regional level. At this time, it is only necessary to determine the regional characteristic information in form. The specific determination steps are as follows:

(1) Determine the number of sub-regions S _Ns+1 of the next region level, if S _Ns+1 >0 go to step (2), otherwise, determine that the current independent geometric figure does not contain region feature information.

(2) Search whether there is a serial number P _Ns + equal to the serial number of the current independent geometric figure in the independent geometric figure serial number P _Ns+1 [i] (1≤i≤S _Ns+1 ) defining the sub-region of the next regional level ₁ [i]=b, if yes, go to step (3), otherwise, determine that the current independent geometric figure does not contain area feature information.

(3) Determine that the current independent geometric figure feature information includes the next area level, the area feature information QTZ _Ns+1 .mk with the serial number mk, where mk=S _c +i.

Step 34: Determine the feature information of the current area, denoted as QTZ _Ns.ms ;

QTZ _Ns.ms is the characteristic information of the area whose area level is N _s and the area sequence number is m _s . QTZ _Ns.ms is expressed as: QTZ _Ns.ms (qpc, Cms, cpc, DTZ _b ,...) the value of serial number b is (g≤b≤g+C _ms -1);

Among them, qpc is the reference point of the current area; C _ms is the number of independent geometric figures contained in the current area; cpc is the reference point of the current area. This vector is called the area orientation vector. When there is no independent geometric figure bounding the area, the vector difference is 0; DTZ _b is the feature information of the independent geometric figure contained in the current area.

Step 35: Add 1 to the serial number of the current area, m _s =m _s +1. If m _s > S _c , it means that the area marked by m _s exceeds the current area level, go to step 36; otherwise, it means that there are sub-areas of the same level in the current area level, and return to step 31;

Step 36: Judging that if m _s >m, it means that the number of the current region exceeds the number of sub-regions in the region where the image object is located, and the process of obtaining region feature information ends, and then go to step 37. Otherwise, add 1 to the marked value of the area level, N _s =N _s +1, and the marked value of the area level points to the area of the next level; at the same time, set the area sequence number discrimination parameter S _c =S _c +S of the N _s area level _Ns , return to step 31;

Step 37: The characteristic information of the image object is constituted by the graded region characteristic information obtained in the above steps, denoted as TTZ. Taking the case where the number of area levels is N, the number of areas is m, and the number of geometric figures is h, the general expression of TTZ is:

QTZ _1.1 (qpc, C ₁ , cpc, DTZ _g1 ... DTZ _e1 ); where g ₁ =1; e ₁ =g ₁ −1+C ₁ ;

QTZ _2.2 (qpc, C ₂ , cpc, DTZ _g2 ... DTZ _e2 ); where g ₂ =e ₁ +1; e ₂ =g ₂ −1+C ₂ ;

·

·

QTZ _Nm (qpc, C _m , cpc, DTZ _gm ... DTZ _em ); where g _m =e _m-1 +1; e _m =g _m -1+C _m ;

In the expression form of TTZ, the general expression form of the feature point information of an independent geometric figure in the feature information of a certain area is: DTZ _b (wl, dTZ _b , QTZ _No.mo ). If the independent geometric figure with the serial number b defines the next-level area, the feature information of the independent geometric figure with the serial number includes the area feature information QTZ _No.mo of the area defined by it at the next level, otherwise it will not include the area feature information. The feature information of independent geometric figures including regional feature information only establishes a formal relationship, and the real feature description is also reflected in the corresponding regional feature information.

In the expression form of TTZ, C ₁ ... C _m is the number of independent geometric figures in m areas. Their cumulative sum is the number h of geometric figures in the area where the image target is located. For the m regions of different levels divided by the region where the image target is located, each region has corresponding region feature information. The arrangement rules of the area serial numbers follow the basic principles of first high level and then low level, and first large area and then small area.

The image object feature information TTZ formed according to the above steps realizes the accurate description of the main features of the image object in layers and regions. But there is still a lack of concise descriptions of the more macroscopic features of image objects. the

In order to achieve a better description effect on the image object, in order to be able to describe the common features of a class of image objects, the basic retrieval information of the image object can be extracted from the feature information of the independent geometric figure and the feature information of the region contained in the image object, and Combining the basic retrieval information of the image object with the region characteristic information constitutes the characteristic information of the image object. the

The basic retrieval information of the image object may include the overall statistical information of all regional characteristic information, and may further include statistical information of each region. In statistical information, you can select necessary statistical information according to actual needs. In this embodiment, according to the general situation, the following statistical information is extracted. The specific process is as follows:

The overall statistical information for the extraction of regional feature information of image targets includes:

The number of regional levels N; the number of regions m; the number of geometric figures h; the ratio K _q of the cumulative sum of the lengths of the boundary curves of geometric figures in each region to the minimum value.

For each area contained in the image target, the statistical information of the extracted area includes:

The maximum value mzc of the boundary curve length of each independent geometric figure in the area; the ratio hzk of the cumulative sum of the boundary curve lengths of all independent geometric figures in the area to mzc; the reference point of each independent geometric figure in the area is the independent geometry with the largest boundary curve length in the area The reference point of the graph is the maximum vector mms and the minimum vector mss in the polar coordinate vector of the pole. the

Regional statistical information is expressed as qjs _i (mzc, hzk, mms, mss), where the value of the regional serial number i is (1≤i≤m).

The basic retrieval information of an image target is denoted as tjs(N, m, h, K _q , qjs ₁ , qjs ₂ , ..., qjs _m ).

The feature information TTZ of the image object is composed of the feature information of the area contained in the image object and the basic retrieval information of the image object. Taking the case where the number of area levels is N, the number of areas is m, and the number of geometric figures is h, the general expression of TTZ is:

TTZ (tjs, QTZ _1.1 , QTZ _2.2 , ..., QTZ _Nm )

tjs(N, m, h, K _q , qjs ₁ , qjs ₂ ,..., qjs _m );

QTZ _1.1 (qpc, C ₁ , cpc, DTZ _g1 ... DTZ _e1 ); where g ₁ =1; e ₁ =g ₁ −1+C ₁ ;

QTZ _2.2 (qpc, C ₂ , cpc, DTZ _g2 ... DTZ _e2 ); where g ₂ =e ₁ +1; e ₂ =g ₂ −1+C ₂ ;

·

·

QTZ _Nm (qpc, C _m , cpc, DTZ _gm ... DTZ _em ); where g _m =e _m-1 +1; e _m =g _m -1+C _m ;

In the present invention, tjs in the feature information TTZ of the image target describes the overall feature of the image target, and the m pieces of regional feature information describe the main features of the image target in layers and regions. the

The overall statistical information of tjs describes the number of regional levels N, the number of regions m, the number of geometric figures h contained in the image target, and the ratio of the maximum value to the minimum value of the cumulative sum of the boundary curve lengths of the geometric figures in each region. Use these data to describe the overall characteristics of the image target statistically. The regional statistical information of tjs in qjs _i includes the maximum value mzc of the boundary curve length of each independent geometric figure in the area, the ratio hzk of the cumulative sum of the boundary curve lengths of all independent geometric figures in the area to mzc, and the reference point of each independent geometric figure in the area. The maximum vector mms and the minimum vector mss in the polar coordinate vector of the polar coordinate vector with the reference point of the independent geometric figure with the maximum boundary curve length in it describe the statistical characteristics of the local image targets in each area. In particular, qjs ₁ describes the overall appearance statistics of image objects.

Among the area feature information, the area feature information QTZ _1.1 with the highest level number 1 describes the main features of the appearance of the entire area of the image object. The feature information of the regions at the following levels describes the main features of the local images of the corresponding regions. The number C _i of independent geometric figures in the area feature information QTZ _ei describes that there are C _i independent geometric figures in the area; for the feature information DTZ _j of C _i independent geometric figures in the area feature information QTZ _ei , its polar coordinates The vector wl describes the mutual positional relationship of C _i independent geometric figures; its feature information dTZ _j of the geometric figures describes the main features of the boundaries of the corresponding independent geometric figures; its area feature information describes the characteristics of the next-level area defined by the independent geometric figures . The feature information of the independent geometric figure included in the feature information of the region, and the feature information of the next-level area contained in the feature information of the independent geometric figure form a relationship describing the main features of the image object in layers and regions.

The inventive method is described below with specific examples:

In this example, the feature information of the image object shown in Figure 4 is obtained and the characteristics of the image object are described, including the following steps:

Step 40: Obtain 6 closed curves by searching for pixels with similar boundary optical parameters, and these 6 closed curves form 6 geometric figures: t1~t6, as shown in Figure 5;

Step 41: Calculate feature information including reference points, a set of feature point information and basic search information for the geometric figures t1-t6 respectively. the

As shown in Figure 6A, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t1, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature point of the geometric figure ;

As shown in Figure 6B, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t2, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature of the geometric figure point;

As shown in Figure 6C, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t3, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature point of the geometric figure ;

As shown in Figure 6D, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t4, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature point of the geometric figure ;

As shown in Figure 6E, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t5, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature point of the geometric figure ;

As shown in Figure 6F, it is a vector schematic diagram of a group of feature point information with the reference point as the pole in the feature information of the geometric figure t6, pc is the reference point of the geometric figure, and the point pointed by the arrow is the feature point of the geometric figure . the

Step 42: Construct feature information of the image object from the obtained feature information of multiple geometric figures, and describe the feature of the image object. the

The image target area shown in FIG. 5 is set as the highest-level area, and the area level is 1. In this area, there are three independent geometric figures, which are recorded as t1, t2, and t3 in order of boundary length. Among them, t2 and t3 are simple geometric figures, the area defined by the boundary of t1 contains independent geometric figure t4, and the area defined by the boundary of independent geometric figure t4 contains geometric figures t5 and t6. the

The image object division shown in FIG. 5 has three levels of regions. The first-level area is the entire image area, including three independent geometric figures t1, t2, and t3; the second-level area is the area defined by the boundary of t1, including an independent geometric figure t4; the third-level area is defined by the boundary of t4 The area contains two independent geometric figures t5 and t6. The number of regions in each level is 1, and the total number of regions is 3. the

According to the method in step 12, the characteristic information of the first-level area, the second-level area and the third-level area are determined. Wherein, the reference point of the first-level area is qpc1, the reference points of t1, t2, and t3 are respectively pc1, pc2, and pc3, and the vector relationship between pc1, pc2, pc3 and qpc1 is shown in Figure 7A; determine the second-level area ( That is, the reference point of the area defined by t1) is qpc2, the reference point of t4 is pc4, and pc4 coincides with qpc2, as shown in Figure 7B; the reference point for determining the third-level area (i.e., the area defined by t4) is qpc3, The reference point of t5 and t6 is pc5, pc6, pc5, and the vector relationship between pc6 and qpc3 is shown in Fig. 7C. the

Finally, according to the method of step 12, the characteristics of the image target are described by the feature information of the first-level area, the second-level area and the third-level area:

QTZ _1.1 (qpc1, C ₁ , cpc ₁ , DTZ ₁ , DTZ ₂ , DTZ ₃ ); wherein, C ₁ takes a value of 3, and cpc ₁ is 0 because there is no independent geometric figure defining the area where the image target is located; DTZ ₁ is the characteristic information of the independent geometric figure t1 contained in this area, DTZ ₂ is the characteristic information of the independent geometric figure t2 contained in this area, and DTZ ₃ is the characteristic information of the independent geometric figure t3 contained in this area;

QTZ _2.2 (qpc2, C ₂ , cpc ₂ , DTZ ₄ ); wherein, C ₂ takes a value of 1, and cpc ₂ is the vector difference between the reference point qpc2 of this area and the reference point pc1 of the independent geometric figure t1 defining this area; DTZ ₄ is the characteristic information of the independent geometric figure t4 contained in this area;

QTZ _3.3 (qpc ₃ , C ₃ , cpc ₃ , DTZ ₅ , DTZ ₆ ; wherein, C ₃ takes a value of 2, cpc ₃ is the reference point qpc3 of this area and the reference point pc4 of the independent geometric figure t4 that bounds this area Vector difference; DTZ ₅ is the characteristic information of the independent geometric figure t5 contained in this area, and DTZ ₆ is the characteristic information of the independent geometric figure t6 contained in this area.

Further, the basic retrieval information tjs(N, m, h, K _q , qjs ₁ , qjs ₂ , qjs ₃ ) of the image target can also be extracted, where N=3; m=3; h=6; K _q =2.64; area Statistical information mzc＝2218 in qjs ₁ (mzc, hzk, mms, mss); hzk＝0.716; mms＝230∠-132; mss＝262∠-135; mzc＝ in qjs ₂ (mzc, hzk, mms, mss) 1160; hzk＝1; mms＝0∠0; mss＝0∠0; mzc＝820 in qjs ₃ (mzc, hzk, mms, mss); hzk＝0.811; mms＝98∠-90; mss＝98∠- 90;

The feature information of the image target is composed of the feature information of the area contained in the image target and the basic retrieval information of the image target:

TTZ (tjs, QTZ _1.1 , QTZ _2.2 , QTZ _3.3 )

tjs(3, 3, 6, 2.64, qjs ₁ , qjs ₂ , qjs ₃ );

QTZ _1.1 (qpc1, ₃ , cpc1, _DTZ1 , _DTZ2 , _DTZ3 );

QTZ _2.2 (qpc2, 1, cpc ₂ , DTZ ₄ );

QTZ _3.3 (qpc3, 2, cpc ₃ , DTZ ₅ , DTZ ₆ );

In this embodiment, tjs in the feature information TTZ of the image object describes the overall features of the image object shown in FIG. 4 , and the three regional feature information describe the main features of the image object shown in FIG. 4 in layers and regions. the

Referring to Figure 8, the embodiment of the present invention also provides a feature description device for an image object, which includes:

Graphics division unit 80, for dividing the image object into a plurality of geometric figures;

A graphic feature determination unit 81, configured to obtain feature information of the geometric figure divided by the graphic dividing unit;

The image feature determining unit 82 is configured to determine feature information of the image object according to a set of geometric figure feature information acquired by the graphic feature determining unit, and describe the feature of the image object. the

The graphic feature determination unit 81 includes:

The curvature calculation unit is used to calculate the boundary curve of the geometric figure obtained according to a certain direction, and obtain the curvature information of each point on the boundary curve;

A reference point obtaining unit is used to obtain the reference point of the boundary curve of the geometric figure;

A feature point obtaining unit, which is used to obtain feature points on the boundary curve of the geometric figure that mark the change feature;

The feature information obtaining unit is configured to calculate feature point information according to the feature point, the reference point and the curvature information, and a set of feature point information obtained from the calculation constitutes feature information of the geometric figure. the

Image feature determination unit 82 includes:

A local unit 821, configured to use the feature information of each geometric figure to describe the local features of the image target;

A change unit 822, configured to use a set of feature point information in the feature information of the geometric figure to describe the change characteristics of the boundary of the geometric figure;

The overall unit 823 is used to describe the overall characteristics of the geometric figure by using the basic retrieval information in the geometric figure feature information;

The optical unit 824 is used to describe the optical characteristics of the geometric figure using the optical parameters in the geometric figure feature information;

The location unit 825 is used to describe the positional relationship between the geometric figures using the reference points in the feature information of the geometric figures;

The feature unit 826 is configured to use a set of feature information of geometric figures to describe the main features of the image object. the

The device further includes an area division unit 83, and the area division unit 83 includes:

The initialization unit is used to set the area where the image target is located as the highest level area, denoted as N ₁ , and set the current area level N _i =N ₁ ;

The area graphic determination unit is used to sequentially determine the independent geometric figures in each sub-area in the sub-areas of the current area level _Ni , and determine the independent geometric figures in each sub-area that define the next-level area, and are not used in the current area. The geometric figures surrounded by other geometric figures are called independent geometric figures, and the independent geometric figures surrounding other geometric figures in the current area can define the next-level area;

Judging unit, used to determine the number S of independent geometric figures bounding the next-level area determined by the area figure determination unit, when S=0 the division process ends, otherwise, the area level is reduced to the next level N _i =N _i+1 , and return to the area graphic determination unit for the current area level graphic determination.

Image feature determination unit 82 includes:

The area reference point calculation unit 827 uses one or more of the reference points of the geometric figures in the area, boundary information, and area information, or uses the area boundary information and/or area information to perform calculations, and the obtained in In the case of rotation, translation and zooming, the point whose relative position remains unchanged relative to the reference point of the independent geometric figure in the said area is used as the reference point of the said area;

The vector difference calculating unit 828 is used to calculate the polar coordinate vector of the reference point of the region with the reference point of the region boundary geometric figure as the pole, and when the region has no boundary geometric figure, the polar coordinate vector is 0 ;

The independent feature obtaining unit 829 is used to obtain the feature information of the independent geometric figure for the independent geometric figure in the region;

The feature forming unit 830 is used to combine the reference point of the region, the polar coordinate vector of the region reference point with the reference point of the region boundary geometric figure as the pole, the number of independent geometric figures in the region and a corresponding one A set of independent geometric figure feature information constitutes the feature information of the region; and the feature information of each divided region is determined as the feature information of the image object. the

Image feature determination unit 82 further includes:

The retrieval information extraction unit 831 is configured to extract various statistical information from the feature information of the image object to form basic image retrieval information describing the overall characteristics of the image object. the

In summary, the beneficial effects of the present invention are:

In the present invention, by dividing the image target into a plurality of geometric figures, obtaining the feature information of each geometric figure, and determining the feature information of the image target according to a group of geometric figure feature information obtained, the process of obtaining the image target feature information is only Computing and processing a small amount of data in the image, compared with the existing technology, effectively reduces the amount of calculation, and because the features of the multiple geometric figures contained in the image target can accurately reflect the characteristics of the image target, so using the image target A set of geometric figure feature information is included to describe the features of the image target, which can effectively improve the accuracy of describing the feature of the image target. the

Secondly, in the solution provided by the embodiment of the present invention, for the plurality of geometric figures constituting the image target, a set of feature point information is used to describe the order of the main features of the boundaries of the geometric figures in a circular ordering relationship. In the case of rotation, translation, and scaling, the unique description of the main features of the geometric figure is realized. On this basis, in the feature information of each area, the polar coordinate vector of the feature information of the independent geometric figures in the area is used to further accurately describe the positional relationship of the independent geometric figures, which solves the problem of the relationship between the feature information of the image target and the image target in the prior art. There is no theoretical one-to-one correspondence between the main features, resulting in the defect of inaccurate description. the

Further, in the solution provided by the embodiment of the present invention, the area where the image target is located is divided into multiple areas of different levels according to the mutual inclusion relationship of multiple geometric figures, and the feature information of each area is obtained, and the multiple area feature information Constitute image target feature information. This constitutes a hierarchical and regional description of the main features of the image target, thereby solving the defect that the image features cannot be accurately described in the prior art. the

Furthermore, in the solution provided by the embodiment of the present invention, the basic retrieval information of the image object is constituted by extracting statistical information reflecting the overall characteristics of the image object from the feature information of the independent geometric figures contained in the image object and the feature information of the region, which describes the image object Statistical features and common features of a class of image objects. It solves the defect that the common features of a class of image objects cannot be described in the prior art. The basic retrieval information of the image target has the functions of improving the recognition speed and retrieving the image information database. the

To sum up, compared with the prior art, the present invention has the remarkable effect of more accurate, more detailed and more comprehensive description of image target features. the

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations. the

Claims (14)

1. the character description method of an image object is characterized in that, this method comprises:

A, in image object the search pixel close with the boundary light mathematic(al) parameter, constitute a plurality of closed curves according to the pixel that searches, with the figure of each closed curve formation as a geometric figure;

B, according to a certain direction the geometric boundary curve that obtains is calculated, and know the curvature information of each point on the boundary curve; Ask for the reference point of described geometric figure border curve; Ask on the curve of described geometric figure border, indicate the unique point of variation characteristic; With described reference point is polar limit, extreme value, the polar angle of calculated characteristics point pole coordinate vector; Calculate acquisition unique point type code, supplementary features code and radius-of-curvature according to curvature information; By extreme value, polar angle and unique point type code, supplementary features code and the radius-of-curvature constitutive characteristic dot information of described polar coordinates vector, and constitute the characteristic information of describing described geometric figure shape according to the characteristic point information that calculates;

One group of geometric figure characteristic information that c, basis are asked for is determined the characteristic information of described image object, and describes the feature of described image object.

2. the method for claim 1 is characterized in that, the described geometric figure characteristic information of asking for further comprises the steps:

Constitute the basic retrieving information of describing the geometric figure global feature by a described geometric boundary information and a described group of feature point information extraction various statistic, basic retrieving information comprises:

The maximum vector of characteristic point information;

The ratio of the maximum vector of characteristic point information and minimum vector extreme value;

The number of all kinds unique point; The sum of unique point;

Characteristic point information vector anglec of rotation absolute value add up and;

The border total length; The ratio of all kinds boundary curve section cumulative length and border total length;

Optical parametric on the curve of geometric figure border;

Constitute the geometric characteristic information of description by described boundary curve reference point, a group of feature point information and basic retrieving information.

3. as arbitrary described method in claim 1 or 2, it is characterized in that the feature of describing described image object among the step c comprises:

Each geometric characteristic information is described the local feature of image object, the variation characteristic on the group of feature point information description geometric figure border in the geometric figure characteristic information, basic retrieving information is described geometric global feature in the geometric figure characteristic information, optical parametric is described geometric optical signature in the geometric figure characteristic information, reference point is described the position relation between each geometric figure in the geometric figure characteristic information, and one group of geometric characteristic information is described the principal character of image object.

4. as arbitrary described method in claim 1 or 2, it is characterized in that, among the step a image object is divided into a plurality of geometric figures after, further the image object region is divided into the subregion of different stage, and the independent geometric figure in definite all subregion, specifically comprise the steps:

The zone at A, described image object place is set to highest zone, is designated as N ₁, the current region level n is set _i=N ₁

B, order are at the current region level n _iSubregion in, determine the independent geometric figure in all subregion, determine to define in all subregion the independent geometric figure in next stage zone, the geometric figure that is not surrounded by other geometric figures in current region is called independent geometric figure, surrounds other geometric independent geometric figure and can define the next stage zone in current region;

The geometric number S of the independence that defines the next stage zone that C, judgement are determined by step B, when the end of S=0 partition process, otherwise regional rank is reduced to next stage N _i=N _I+1, change step B.

5. method as claimed in claim 4 is characterized in that, described step b further asks for area characteristic information by the zone after asking for each geometric characteristic information, comprises the steps:

D, utilize in the described zone one or more in geometric reference point, boundary information, the area information, perhaps utilize described zone boundary information and/or area information to calculate, obtain in rotation, under the Pan and Zoom situation, with the constant point of independent geometric figure reference point relative position in the described zone, as the reference point in described zone;

E, to ask described regional reference point be the polar coordinates vector of limit with described zone boundary geometric figure reference point, and when described regional non-boundary geometric figure, described polar coordinates vector is 0;

F, to the independent geometric figure in the described zone, ask independent geometric figure characteristic information;

G, be that independent geometric number and one group of corresponding independence geometric figure characteristic information constitute area characteristic information in the polar coordinates vector of limit, the described zone with described zone boundary geometric figure reference point by the reference point in described zone, described regional reference point.

6. method as claimed in claim 5 is characterized in that, the described geometric characteristic information of independence of asking for comprises:

Asking in the described zone independent geometric reference point is the polar coordinates vector of limit with the reference point in described zone;

By the geometric characteristic information of the polar coordinates vector of asking for, described independent geometric figure correspondence and the area characteristic information of described independent geometric figure delimited area, constitute the geometric characteristic information of described independence.

7. method as claimed in claim 6 is characterized in that, the feature of describing described image object among the step c comprises:

Describe the profile principal character of described image object with the characteristic information in highest zone, describe the principal character of corresponding region topography target with the characteristic information in following zones at different levels;

In area characteristic information, describing this zone with independent geometric number n has n independent geometric figure; With regional reference point and described polar coordinates vector, this regional relative position relation is described; With the polar coordinates vector in n the independent geometric characteristic information n independent geometric figure mutual alignment relation described, describe the principal character on corresponding independent geometric figure border with geometric figure characteristic information corresponding in n the independent geometric figure characteristic information, describe the principal character in the next stage zone that corresponding independent geometric figure defines with the area characteristic information in the individual independent geometric figure characteristic information of n;

Independent geometric figure characteristic information in the area characteristic information inclusion region, independent geometric figure characteristic information comprises the area characteristic information of next stage;

By area characteristic information and independent geometric figure characteristic information constitutes by different level, described image object principal character is described in the subregion relation.

8. the method for claim 1 is characterized in that, extracts various statistic by described independent geometric figure characteristic information and area characteristic information and constitutes the basic retrieving information of image object of describing described image object global feature.

9. method as claimed in claim 8 is characterized in that, the characteristic information of described image object is made of the characteristic information and the basic retrieving information of described image object in the zones at different levels that mark off.

10. the characterization device of an image object is characterized in that, this device comprises:

The figure division unit is used for constituting a plurality of closed curves in the image object search pixel close with the boundary light mathematic(al) parameter according to the pixel that searches, with the figure of each closed curve formation as a geometric figure;

The graphic feature determining unit comprises that curvature asks for that the unit is asked in unit, reference point, unique point asks for the unit and characteristic information is asked for the unit; Described curvature is asked for the unit, is used for according to a certain direction the geometric boundary curve that obtains being calculated, and knows the curvature information of each point on the boundary curve; Described reference point is asked for the unit, is used to ask for the reference point of described geometric boundary curve; Described unique point is asked for the unit, is used to ask for described geometric boundary curve, indicates the unique point of variation characteristic; Described characteristic information is asked for the unit, and being used for described reference point is polar limit, extreme value, the polar angle of calculated characteristics point pole coordinate vector; Calculate acquisition unique point type code, supplementary features code and radius-of-curvature according to curvature information; By extreme value, polar angle and unique point type code, supplementary features code and the radius-of-curvature constitutive characteristic dot information of described polar coordinates vector, the characteristic point information that calculates constitutes the characteristic information of describing described geometric figure shape;

The characteristics of image determining unit is used for determining the characteristic information of described image object according to one group of geometric figure characteristic information that described graphic feature determining unit is asked for, and describes the feature of described image object.

11. device as claimed in claim 10 is characterized in that, described characteristics of image determining unit comprises:

Local unit is used to use each geometric characteristic information to describe the local feature of image object;

Change unit is used for using the variation characteristic on a group of feature point information description geometric figure border of geometric figure characteristic information;

Integral unit is used for using the basic retrieving information of geometric figure characteristic information to describe geometric global feature;

Optical unit is used for using geometric figure characteristic information optical parametric to describe geometric optical signature;

Position units, the position that is used for using geometric figure characteristic information reference point to describe between each geometric figure concerns;

Feature unit is used to use one group of geometric characteristic information to describe the principal character of image object.

12., it is characterized in that this device further comprises as arbitrary described device in the claim 10:

Initialization unit, the zone that is used for described image object place is set to highest zone, is designated as N ₁, the current region level n is set _i=N ₁

The regional graphics determining unit is used for order at the current region level n _iSubregion in, determine the independent geometric figure in all subregion, determine to define in all subregion the independent geometric figure in next stage zone, the geometric figure that is not surrounded by other geometric figures in current region is called independent geometric figure, surrounds other geometric independent geometric figure and can define the next stage zone in current region;

Judging unit is used to judge the geometric number S of the independence that defines the next stage zone that is determined by described regional graphics determining unit, and when the end of S=0 partition process, otherwise regional rank is reduced to next stage N _i=N _I+1, and return area figure determining unit is carried out other figure of current region level and is determined.

13. device as claimed in claim 12 is characterized in that, described characteristics of image determining unit comprises:

The zone reference point is asked for the unit, be used for utilizing the one or more of geometric reference point in the described zone, boundary information, area information, perhaps utilize described zone boundary information and/or area information to calculate, obtain in rotation, under the Pan and Zoom situation, with the constant point of independent geometric figure reference point relative position in the described zone, as the reference point in described zone;

Phasor difference is asked for the unit, and the reference point that is used to ask for described zone is the polar coordinates vector of limit with described zone boundary geometric figure reference point, and when described regional non-boundary geometric figure, described polar coordinates vector is 0;

Independent characteristic is asked for the unit, is used for the independent geometric figure in the described zone is asked for the geometric characteristic information of this independence;

Feature constitutes the unit, and being used for the reference point in described zone, described regional reference point is the characteristic information that independent geometric number and one group of corresponding independence geometric figure characteristic information constitute described zone in the polar coordinates vector of limit, the described zone with described zone boundary geometric figure reference point.

14. device as claimed in claim 13 is characterized in that, described characteristics of image determining unit further comprises:

The retrieving information extraction unit is used for extracting various statistic by geometric characteristic information of described independence and area characteristic information and constitutes the basic retrieving information of image of describing described image object global feature.

Images