CN101833750A - Active Contouring Method and System Based on Shape Constraint and Orientation Field - Google Patents

Abstract

The invention discloses an active contour method based on shape constraint and direction field, and a system thereof. The method comprises the following steps of: S1, inputting the initial contour of a target object in image, and establishing the energy field relating to the shape distance of the initial contour according to the shape priori knowledge of the initial contour; S2, establishing the gradient intensity energy field by calculating the gradient intensity values extracted from all points of the edge of the target object; S3, establishing the gradient direction energy field by calculating the consistency between the tangential direction of each point in the initial contour and the tangential direction of each point extracted from the edge of the target object; and S4, overlapping the three energy fields established in S1-S3 to obtain a new energy field, carrying out global optimization, and obtaining an accurate contour of the target object. The invention gives more accurate and complete expression of target image contour under the condition of target image geometrical information in the known image, thus not only completely maintaining the geometrical characteristics of the target image, but also enriching the contour details.

Description

Active Contouring Method and System Based on Shape Constraint and Orientation Field

technical field

The invention belongs to the technical field of image processing, in particular to an active contour method and system based on shape constraints and direction fields.

Background technique

In recent years, with the development of computer hardware and software technology and the popularization of a large number of image and video equipment, computer vision technology has become a hot field of computer research. Among them, the detection, extraction and processing of specific objects in image and video data have high practical value and technical significance. The object detection methods provided by the current vision technology are usually aimed at texture and regional features, such as the Haar feature method for face detection, the signal processing method for flame detection, etc., which often can only provide the processor with a specific object position and The rough estimation of the area, but the more accurate estimation and extraction of the edge of a specific object cannot be performed. Therefore, in order to solve the problem of edge extraction, the traditional active contour method (Active contour) method came into being. The main idea of the traditional active contour method is to first input an initial contour, and then use the information of the energy field (energy is calculated by gradient or other features) input from the outside to find the best position of each contour point through optimization. To find a more accurate boundary. In order to improve the traditional method, the Active contour with shape prior method (Active contour with shape prior) appeared later, which can find the optimal boundary while maintaining the original shape. However, this method does not make full use of the actual information of the image—especially the meaningful boundary information of each primitive in the image—and it is easy to cause the loss of boundary details.

Therefore, the active contour method technology should be developed towards the direction of extracting the object boundary more accurately, while meeting the needs of some constraints.

Contents of the invention

(1) Technical problems to be solved

The technical problem to be solved by the present invention is how to accurately extract the edge of the image under the condition of satisfying the shape constraint without causing loss of edge details.

(2) Technical solution

In order to solve the above problems, the present invention provides an active contour method based on shape constraints and direction fields, comprising the following steps:

S1, input the initial contour of the target object in the image, and establish an energy field related to the shape distance of the initial contour according to the prior knowledge of the shape of the initial contour;

S2, by calculating the value of the gradient strength of each point of the edge extracted in the target object to establish a gradient strength energy field;

S3, by calculating the consistency of the tangent direction of each point in the initial contour and the tangent direction of each point of the edge extracted in the target object, to establish a gradient direction energy field;

S4, superimpose the three energy fields established in steps S1-S3 to obtain a new energy field, and then perform global optimization on it to obtain an accurate outline of the target object in the image.

Among them, the shape prior knowledge of the initial contour can indicate the geometric shape, location and approximate area occupied by the target object.

Among them, in step S1, the step of establishing the energy field related to the shape distance of the initial contour is specifically: according to the position of each point of the initial contour, calculate the corresponding point between each point and the point along the normal direction of the initial contour , and then calculate the normalized energy according to the distance, and obtain the value of the energy field related to the original shape distance of the initial contour.

Wherein, in step S3, by calculating the gradient direction of each point of the edge extracted in the target object, the tangent direction of each point of the edge extracted in the target object is obtained.

Among them, in step S4, according to the elastic coefficient and stiffness coefficient of each point on the initial contour, use an iterative method to find the optimal position of each point on the initial contour within its own neighborhood until the algorithm converges, and obtain the accurate contour of the target object .

Wherein, in step S3, after normalizing the vector of the tangent direction of each point in the initial contour and the vector of the tangent direction of each point of the edge extracted in the target object, the dot product of the two is regarded as the consistent The measure of the property, the dot product is the value of the energy field in the gradient direction.

The present invention also provides an active contour system based on shape constraints and direction fields, including:

The energy field calculation unit related to the original shape distance is used to input the initial contour of the target object in the image, and according to the shape prior knowledge of the initial contour, establish the energy field related to the shape distance of the initial contour;

a gradient strength energy field calculation unit, configured to establish a gradient strength energy field by calculating the value of the gradient strength of each point of the edge extracted in the target object;

The gradient direction energy field calculation unit is used to establish the gradient direction energy field by calculating the consistency of the tangent direction of each point in the initial contour and the tangent direction of each point of the edge extracted in the target object;

The global optimization unit is used to superimpose the three energy fields established by the above three units to obtain a new energy field, and then perform global optimization on it to obtain an accurate outline of the target object in the image.

(3) Beneficial effects

The technical solution of the present invention not only considers the restriction of the shape prior knowledge on the resulting contour, but also makes the contour closer to the original visible edge through the information of the direction field (embodied in the establishment of the gradient strength energy field and the gradient direction energy field), This makes the extracted contour more accurate. In short, this method can give a more accurate and complete expression of the contour of the target image when the geometric information of the target image in the image is known, not only completely retaining the geometric features of the target image, but also enriching its contour details .

Description of drawings

Fig. 1 is the method flowchart of the embodiment of the present invention;

Fig. 2 is three kinds of energy diagrams obtained by implementing the method of the embodiment of the present invention;

Fig. 3 is a diagram of the results of implementing the method of the embodiment of the present invention.

Detailed ways

The specific implementation manners of the present invention will be described in further detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

Fig. 1 is a flow chart of the method of the embodiment of the present invention. As shown in Figure 1, an initial outline of the target image is first input, which can correctly indicate the geometric shape of the target image, and can mark the location and approximate area occupied by the target image. Each point on the initial contour is stored clockwise, and each point is recorded as {X _i }, and at the same time, the tangent direction at each point of the initial contour is calculated, which is recorded as {dX _i }.

Next, three kinds of energy diagrams need to be established (as shown in Figure 2):

First, according to the position of each point on the initial contour, find out which point on the contour corresponds to each point on the image along the corresponding contour normal direction, calculate its distance [D(V) _i ], and obtain the energy related to the distance of the original shape field (that is, the shape prior energy field), and its energy value (the value of the energy field related to the original shape distance) is:

E _{shape, i} = N((1+D(V) _i ) ^-1

Wherein, N represents a normalization function, so that the energy value is between 0 and 1. This energy is used as the influence mode of the shape constraint on the contour change in the present invention. It can be seen that the closer to the original position, the greater the energy, and the higher the reliability of the match.

Then, edge extraction is performed on the image. The present invention adopts the method for calculating the gradient to extract the edge, and the intensity value of the edge is the value of the said gradient intensity energy field, which is recorded as:

E _{strength, i} = S(V _i )

At the same time, according to the extracted edge gradient direction, the edge tangent direction at each point is calculated. According to the correspondence relationship found above, after normalizing the direction vectors of the corresponding contour points and points on the image, the two The dot product serves as a measure of its directional consistency, namely:

E _{direction, i} = N(<dV _i , dX _i >)

Finally, the above three energy values are superimposed to obtain the optimized energy field required to adjust the profile in the present invention:

E _i = αE _{shape, i} + βE _{direction, i} + γE _{strength, i}

而α、β、γ的值可以根据所要提取的目标对象性质的不同进行选择：若目标对象被遮挡严重，则取α为大于0.5的值；若初始轮廓与目标对象比较接近，则α为小于0.5的值；并且一般取β＝γ。The coefficients for the above energy superposition need to satisfy

The values of α, β, and γ can be selected according to the nature of the target object to be extracted: if the target object is seriously occluded, take α as a value greater than 0.5; if the initial contour is relatively close to the target object, then α is less than 0.5; and generally take β=γ.

After the optimized energy field is established, in order to ensure the smoothness and continuity of the contour, the coefficient of elasticity and the coefficient of stiffness are introduced—represented by the first-order derivative and second-order derivative of the initial contour curve at this point, and each point is at Find the optimal position within its own neighborhood and iterate until the algorithm converges. The result obtained at this time is considered to be the accurate outline of the target image. As shown in Figure 3, the thick solid outline shows the results when the algorithm converges.

The present invention also provides an active contour system based on shape constraints and direction fields, including:

The energy field calculation unit related to the original shape distance is used to input the initial contour of the target object in the image, and according to the shape prior knowledge of the initial contour, establish the energy field related to the shape distance of the initial contour;

a gradient strength energy field calculation unit, configured to establish a gradient strength energy field by calculating the value of the gradient strength of each point of the edge extracted in the target object;

The gradient direction energy field calculation unit is used to establish the gradient direction energy field by calculating the consistency of the tangent direction of each point in the initial contour and the tangent direction of each point of the edge extracted in the target object;

The global optimization unit is used to superimpose the three energy fields established by the above three units to obtain a new energy field, and then perform global optimization on it to obtain an accurate outline of the target object in the image.

It can be seen from the above embodiments that this technical solution not only considers the constraints of shape prior knowledge on the resulting contour, but also makes the contour closer to The original visible edge makes the extracted contour more accurate. In short, this method can give a more accurate and complete expression of the contour of the target image when the geometric information of the target image in the image is known, not only completely retaining the geometric features of the target image, but also enriching its contour details .

The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, some improvements and modifications can also be made, these improvements and modifications It should also be regarded as the protection scope of the present invention.

Claims (7)

1. the driving wheel contour method based on the shape constraining and the field of direction is characterized in that, may further comprise the steps:

S1, the initial profile of destination object in the input picture according to the shape prior knowledge of this initial profile, is set up the relevant energy field of shape distance with this initial profile;

S2, the value of the gradient intensity of the every bit by calculating the edge that is extracted in the destination object is set up the gradient intensity energy field;

S3, the consistance of the tangential direction of the every bit at the edge that is extracted in tangential direction by calculating every bit in the initial profile and the destination object is set up the gradient direction energy field;

S4, three energy field stacks that step S1～S3 is set up obtain new energy field, then it are carried out global optimization, obtain the accurate profile of destination object in the image.

2. the driving wheel contour method based on the shape constraining and the field of direction as claimed in claim 1 is characterized in that, geometric configuration, present position and occupied general area that the shape prior knowledge of this initial profile can the indicating target object.

3. the driving wheel contour method based on the shape constraining and the field of direction as claimed in claim 2, it is characterized in that, in step S1, set up with the shape of this initial profile step and be specially: according to the position of each point of initial profile apart from relevant energy field, calculate the distance of every bit and this corresponding point on the initial profile normal direction, go out normalized energy according to this distance calculation then, obtain the value of the energy field relevant with the original-shape distance of initial profile.

4. the driving wheel contour method based on the shape constraining and the field of direction as claimed in claim 2, it is characterized in that, in step S3, the gradient direction of the every bit by calculating the edge that is extracted in the destination object obtains the tangential direction of the every bit at the edge that extracted in the described destination object.

5. the driving wheel contour method based on the shape constraining and the field of direction as claimed in claim 2, it is characterized in that, in step S4, elasticity coefficient and hard lineal number according to each point on the initial profile, utilize process of iteration to seek on the initial profile each optimal location in self neighborhood scope, until algorithm convergence, obtain the accurate profile of destination object.

6. as claim 2 or 4 described driving wheel contour methods based on the shape constraining and the field of direction, it is characterized in that, in step S3, after the normalization of the vector of the tangential direction of the every bit at the edge that is extracted in the vector of the tangential direction of every bit in the initial profile and destination object difference, as described conforming tolerance, this dot product is the value of described gradient direction energy field with the two dot product.

7. the active profile system based on the shape constraining and the field of direction is characterized in that, comprising:

With the relevant energy field computing unit of original-shape distance, be used for the initial profile of input picture destination object, according to the shape prior knowledge of this initial profile, set up the relevant energy field of shape distance with this initial profile;

Gradient intensity energy field computing unit, the value that is used for the gradient intensity of the every bit by calculating the edge that destination object extracted is set up the gradient intensity energy field;

Gradient direction energy field computing unit is used for the consistance of tangential direction of the every bit at the edge that extracted in tangential direction by calculating the initial profile every bit and the destination object, sets up the gradient direction energy field;

The global optimization unit, three energy field stacks that are used for above three unit are set up obtain new energy field, then it are carried out global optimization, obtain the accurate profile of destination object in the image.

Images