CN107230194A - A kind of smooth filtering method based on object point set - Google Patents

Abstract

The invention discloses a smoothing filtering method based on an object point set, comprising the steps of: obtaining a boundary 1; obtaining a boundary 2; performing smoothing and filtering processing on the area point set between the boundary 1 and the boundary 2, and the boundary 1 is an object point set is the edge point set after 10 times of erosion, and the boundary 2 is the edge point set of the object point set after 10 times of expansion. The method created by the present invention utilizes the method of erosion and expansion to obtain boundary 1 and boundary 2, and utilizes boundary 1 and boundary 2 to obtain an area point set, which reflects the outline information of the entire object point set, and performs a process on the area point set Smoothing filter processing can eliminate block effects between objects on the picture and between objects and the background, and improve the overall quality of the picture.

Description

A Smoothing Filtering Method Based on Object Point Set

technical field

The invention relates to the technical field of image processing, in particular to a smoothing filtering method based on an object point set.

Background technique

The latest object extraction technology Mask R-CNN, this technology performs semantic segmentation on the image, and the set of pixels of the same category obtained by segmentation is called the object point set. A picture includes several object point sets. If a certain object point set is processed, such as adjusting the contrast of a certain object point set, after the adjustment, the distance between the object point set and the object point set, the object point Block effects are prone to appear between the set and the background, and these block effects will affect the quality of the entire picture. Therefore, how to eliminate the block effect between the object point set and the object point set, and between the object point set and the background has become an urgent problem in the field.

Contents of the invention

In order to solve the above problems, the present invention provides a smoothing filtering method based on object point sets, which can effectively solve the block effect between object point sets and object point sets, and between object point sets and background.

The solution of the present invention to solve its technical problem is: a kind of smooth filtering method based on the object point set, comprising steps: obtain boundary 1; obtain boundary 2; carry out smooth filter processing to the area point set between boundary 1 and boundary 2, The boundary 1 is the edge point set of the object point set after 10 erosions, and the boundary 2 is the edge point set of the object point set after 10 expansions.

Further, the method for said 10 times of corrosion comprises steps:

a1) Subtracting the current edge point set from the current object point set to obtain a new object point set;

a2) Obtaining a new edge point set with the new object point set;

a3) Repeat steps a1)-a2) 10 times.

Further, the method for obtaining the edge point set includes: taking any pixel point in the object point set as the center, which is called a central pixel point, and judging the relationship between the left, right, upper, and lower pixel points of the central pixel point Whether they belong to the same category, the edge points are obtained according to the judgment result, and the edge points are collected into an edge point set.

Further, the method of said 10 expansions comprises steps:

a11) Put the center of the structure element on any point of the current edge point set, and move the center of the structure element along the edge point set, find the pixel point that belongs to the structure element but does not belong to the object point set, and put the The set of pixel points is a new set of edge points;

a12) merging the current object point set and the edge point set into a new object point set;

a13) Repeat steps a11)-a12) 10 times.

Further, a 3×3 filtering template is used to perform smoothing filtering on the region point set between boundary 1 and boundary 2.

The beneficial effect of the present invention is: this method utilizes the method of erosion and dilation to obtain boundary 1 and boundary 2, and utilizes boundary 1 and boundary 2 to obtain the area point set, and this area point set reflects the outline information of the whole object point set, for this The area point set is processed by smoothing filter, which can eliminate the block effect between objects on the picture and between objects and the background, and improve the overall quality of the picture.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly describe the drawings that need to be used in the description of the embodiments. Apparently, the described drawings are only some embodiments of the present invention, not all embodiments, and those skilled in the art can obtain other designs and drawings based on these drawings without creative work.

Fig. 1 is the flow chart of the smoothing filter method based on object point set;

Fig. 2 is the flowchart of 10 corrosions;

Figure 3 is a flowchart of 10 expansions.

detailed description

The idea, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, features and effects of the present invention. Apparently, the described embodiments are only some of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative efforts belong to The protection scope of the present invention. In addition, all the connection/connection relationships mentioned in this article do not refer to the direct connection of components, but mean that a better connection structure can be formed by adding or reducing connection accessories according to specific implementation conditions. The various technical features in the invention can be combined interactively on the premise of not conflicting with each other.

Embodiment 1, with reference to Fig. 1, the picture processed by Mask R-CNN technology is processed as follows: a smoothing filter method based on object point set,

S00: find the object to be processed and obtain its current object point set;

S01: The object point set undergoes 10 erosions to obtain boundary 1;

S02: The object point set undergoes 10 expansions to obtain boundary 2;

S03: Perform smoothing and filtering processing on the area point set between the boundary 1 and the boundary 2.

It should be noted that the sequence of steps S01 and S02 is not limited, and this implementation takes step S01 first. With reference to Fig. 2, wherein, the method for 10 corrosions comprises the steps:

S1: Subtract the current edge point set from the current object point set to obtain a new object point set;

S2: Obtain a new edge point set by using the new object point set;

S3: Repeat steps S1-S2 10 times.

The method of obtaining the edge point set includes: assuming that the object point set is ∑( _xi , y _i ), in the label category map, each pixel point ( _xi , y _i ) with ∑ ( _xi , y _i ) ) as the center, pixel points ( _xi -1,y _i ), ( _xi +1,y _i ) are the left and right points of the pixel point ( _xi ,y _i ) respectively; ( _xi ,y _i+1 ) ,(x _i , y _i -1) are the upper and lower points of point (x _i , y _i ), respectively. If the pixel ( _xi , y ₎ satisfies the following formula:

|C(x _i -1,y _i )-C(x _i ,y _i )|+|C(x _i ,y _i )-C(x _i+ 1,y _i )|≥1

Or |C(x _i ,y _i+1 )-C(x _i ,y _i )|+|C(x _i ,y _i )-C(x _i ,y _i-1 )|≥1, then determine the The pixel point ( _xi , y _i ) is an edge point, and the edge point is collected into an edge point set. The interpretation of this formula is: the change of the category value of the pixel point label of the pixel point ( _xi , y _i ) from left to right or from top to bottom is greater than or equal to 1. Then it is determined that the pixel point ( _xi , y _i ) is an edge point, where the function C(x, y) refers to the label category value of the pixel point (x, y).

Referring to Figure 3, the method of 10 expansions includes steps:

S11: Put the center of the structure element on any point of the current edge point set, and move the center of the structure element clockwise along the edge point set, find the pixel that belongs to the structure element but not the object point set, and set The set of pixel points is a new set of edge points; expressed as: Among them, A is the current edge point set, B is the structural element, and z is the edge point set obtained by one expansion.

S12: Merge the current object point set and the edge point set into a new object point set;

S13: Repeat steps S11-S12 10 times.

After boundary 1 and boundary 2 are determined, the area point set between boundary 1 and boundary 2 is obtained, and the area point set satisfies the following conditions:

Let the pixel point (x, y) represent any point on the area point set, the pixel point (x1, y1) represent any point on the boundary 1, and the pixel point (x2, y2) represent any point on the boundary 2, then satisfy x1=x= When x2, y1<y<y2 or y2<y<y2.

Perform smoothing and filtering processing on the regional point set between boundary 1 and boundary 2. This embodiment uses a 3×3 filtering template, which is:

The smoothing filtering process is as follows:

Suppose the following is the gray value matrix of an image:

0 0 143 0 0 0 93 157 89 0 189 156 162 0 56 0 89 0 0 0 0 0 0 0 0

Gray value f(3,1)=189, f(3,2)=156, f(3,3)=162 after template filtering

The area point set between boundary 1 and boundary 2 is converted into a gray value matrix, and each pixel is sequentially smoothed and filtered according to the above method to obtain the final smoothed area point set.

The method created by the present invention utilizes the method of erosion and expansion to obtain boundary 1 and boundary 2, and utilizes boundary 1 and boundary 2 to obtain an area point set, which reflects the outline information of the entire object point set, and performs a process on the area point set Smoothing filter processing can eliminate block effects between objects on the picture and between objects and the background, and improve the overall quality of the picture.

The preferred embodiments of the present invention have been described in detail above, but the invention is not limited to the described embodiments, and those skilled in the art can also make various equivalent modifications or replacements without violating the spirit of the present invention. , these equivalent modifications or replacements are all included within the scope defined by the claims of the present application.

Claims (5)

1. A smoothing filter method based on object point set, it is characterized in that: comprise steps: obtain boundary 1; Obtain boundary 2; The area point set between boundary 1 and boundary 2 is carried out smoothing filter process, described boundary 1 is The edge point set of the object point set undergoes 10 times of erosion, and the boundary 2 is the edge point set of the object point set of 10 times of expansion. 2. a kind of smoothing filtering method based on object point set according to claim 1, is characterized in that: the method for described 10 erosions comprises steps: a1) Subtracting the current edge point set from the current object point set to obtain a new object point set; a2) Obtaining a new edge point set with the new object point set; a3) Repeat steps a1)-a2) 10 times. 3. A kind of smoothing filter method based on object point set according to claim 2, it is characterized in that: the acquisition method of described edge point set comprises: take any pixel point in object point set as the center, this point is called For the central pixel point, judge whether the left, right, upper, and lower pixel points of the central pixel point belong to the same category, obtain edge points according to the judgment result, and gather the edge points into an edge point set. 4. a kind of smoothing filter method based on object point set according to claim 1, is characterized in that: the method for described 10 expansions comprises steps: a11) Put the center of the structure element on any point of the current edge point set, and move the center of the structure element along the edge point set, find the pixel point that belongs to the structure element but does not belong to the object point set, and put the The set of pixel points is a new set of edge points; a12) merging the current object point set and the edge point set into a new object point set; a13) Repeat steps a11)-a12) 10 times. 5. A smoothing filtering method based on an object point set according to claim 1, wherein a 3×3 filtering template is used to perform smoothing filtering on the area point set between boundary 1 and boundary 2.