CN115829943A - Image difference region detection method based on super-pixel segmentation - Google Patents
Image difference region detection method based on super-pixel segmentation Download PDFInfo
- Publication number
- CN115829943A CN115829943A CN202211432673.5A CN202211432673A CN115829943A CN 115829943 A CN115829943 A CN 115829943A CN 202211432673 A CN202211432673 A CN 202211432673A CN 115829943 A CN115829943 A CN 115829943A
- Authority
- CN
- China
- Prior art keywords
- image
- difference
- information
- pyramid
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an image difference region detection method based on superpixel segmentation, which comprises the following steps of S1, registering images to be compared by using a sift algorithm, changing the images to be compared to the same plane, and corresponding the coordinates of corresponding points one by one; s2, constructing respective image pyramids of the source image and the target image through a Gaussian function; s3, extracting a first image at the topmost layer of the target image pyramid, performing SLIC processing on the first image, and dividing the image into a plurality of sub-regions with relatively consistent information; s4, differentiating and distinguishing each subregion: the magnitude, direction and color information of the gradient of the pixels of each sub-region are calculated, and the 3 dimensional information is compared with the corresponding pixels in difference. The method utilizes the SLIC algorithm to perform region clustering on the images, and uses the clustered regions as image blocks to perform difference calculation so as to make the information in the blocks more consistent.
Description
Technical Field
The invention belongs to the field of image discrimination, and relates to an image difference region detection method based on superpixel segmentation.
Background
The image discrimination is used for calculating the difference region between the images, provides more useful information in a time sequence space, and has important influence in the fields of detection, tracking and the like.
Some image discrimination algorithms calculate the difference between pixels by taking the pixels as targets; some are directed to block-wise computation of differences between image blocks. The two modes have advantages and disadvantages respectively, the calculation taking the pixel as a target is more precise, and the micro change can be better reflected; the calculation with the block as the target is more robust, and the anti-interference capability on the influence of noise and the like is stronger; meanwhile, the information calculated based on the pixels is more noisy and is easily influenced by factors such as noise points, offset and the like; based on block calculation, the difference of information is larger due to the fact that the components in the region are inconsistent, and useful information is filtered out by screening the information, so that the detection of the difference region is not fine and smooth enough.
Disclosure of Invention
In order to solve the above problems, the present invention provides a method for detecting an image difference region based on superpixel segmentation, which comprises the following steps:
s1, registering images to be compared by using a sift algorithm, changing the images to be compared to the same plane, and corresponding coordinates of corresponding points one by one;
s2, constructing respective image pyramids of the source image and the target image through a Gaussian function;
s3, extracting a first image at the topmost layer of the target image pyramid, performing SLIC processing on the first image, and dividing the image into a plurality of sub-regions with relatively consistent information;
s4, differentiating and distinguishing each subregion: calculating the size, the direction and the color information of the gradient of the pixel of each subarea, performing difference comparison on the 3 pieces of dimension information and the corresponding pixels, and adding 1 to the difference of the point when the difference value of a certain dimension is greater than a preset threshold value; when the information difference values of 3 dimensions are all larger than the threshold value, the difference degree of the point is additionally added with 1, namely the difference value of each pixel point is between 0 and 4.
Preferably, the image pyramid in S2 includes two parameters, namely, a layer height and a layer number, where the layer number determines the resolution of how many kinds of images are in the pyramid, the layer height determines the number of the filtered images at a single resolution, and the shape of the pyramid is adjusted by the two parameters, so that the ratio of the detection area at the difference to the total detection area in the subsequent processing is the largest.
Preferably, after S3, the method further includes using an interpolation algorithm to expand the segmentation information of the first image at the top level of the pyramid onto each layer of the pyramid, so that all images of the pyramid obtain corresponding and consistent image area distribution.
Preferably, the S4 includes the steps of:
s41, calculating the gradient amplitude and direction of the pixel by using a sobel operator:
I x ＝G x I (1)
I y ＝G y I (2)
wherein I is an input image, G x And G y Sobel operators in x-and y-directions, respectively, I x And I y As a gradient map in the corresponding direction, I I Is the magnitude of the gradient, I θ Is the direction of the gradient;
s42, converting the image into an HSV color space, and acquiring the value of the H dimension, namely the color information I of the pixel H Respectively obtaining I of source image pixel and target image pixel G 、I θ And I H After 3 dimensions of information, calculating difference values corresponding to 3 dimensions:
wherein, I' G 、I′ θ 、I′ H Is the corresponding 3-dimensional information value of the source image,
is the corresponding 3-dimensional information value of the target image when S G 、S θ 、S H Respectively greater than a gradient threshold tau G 、τ θ 、τ H The total difference D of the pixels p Respectively adding 1, otherwise, when S G 、S θ 、S H When all three are greater than the corresponding threshold value, D p Then additionally adding 1;
s43, summing and normalizing the difference degrees of all pixels in the region to obtain the total difference degree D of the region t ：
Wherein n is the number of regional pixels, D pi Is the ith pixel D p Value of, when the area is totally different by degree D t Greater than a threshold τ t If so, marking the area as 1, namely, a difference area possibly exists, otherwise, recording the table as 0; after marking all image areas in the pyramid, when the number of images marked as 1 in a certain area is larger than the total number of images tau r If so, determining that the area has a difference;
in formula (9), M k The label value for the region that is desired to be indexed by k for the final disparity map, m is the total number of pyramid images,
for the ith imageThe label values of k regions.
The beneficial effects of the invention at least comprise:
based on the detection algorithm of the image block, but different from the simple division of the image, the method and the device utilize the SLIC algorithm to perform area clustering on the image, and perform difference calculation by taking the clustered area as the image block, so that the information in the block is more consistent. Meanwhile, the information in the block is calculated pixel by pixel, the gradient module and angle information are combined, and the hue information is used for supplementing, compared with a histogram quantization mode, the pixel-by-pixel calculation method can not discard the spatial information of the pixels, so that the calculation result is more accurate and reliable.
Drawings
FIG. 1 is a flowchart illustrating the steps of a method for detecting a difference region in an image based on superpixel segmentation according to an embodiment of the present invention;
FIG. 2 is a pyramid image of an image in the method for detecting image difference regions based on superpixel segmentation according to the embodiment of the present invention;
FIG. 3 is a clustering input diagram of an image difference region detection method based on superpixel segmentation according to an embodiment of the present invention;
FIG. 4 is a graph of the cluster segmentation effect of the super-pixel segmentation based image difference region detection method according to the embodiment of the present invention;
FIG. 5 is a diagram illustrating the difference detection effect of the method for detecting the difference region of an image based on superpixel segmentation according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
On the contrary, the invention is intended to cover alternatives, modifications, equivalents and alternatives which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of the present invention, certain specific details are set forth in order to provide a better understanding of the present invention. It will be apparent to one skilled in the art that the present invention may be practiced without these specific details.
Referring to fig. 1, the method comprises the following steps:
s1, registering images to be compared by using a sift algorithm, enabling the images to be changed to be under the same plane, and enabling coordinates of corresponding points to correspond one to one, so that influences caused by different shooting poses are eliminated;
s2, constructing respective image pyramids of the source image and the target image through a Gaussian function, and referring to FIG. 2;
s3, extracting a first image at the topmost layer of the target image pyramid, performing SLIC (simple linear iterative clustering) processing on the first image, dividing the image into a plurality of sub-regions with relatively consistent information, and improving the operation speed and clustering effect on the k-means algorithm by the SLIC (simple linear iterative clustering);
referring to fig. 3 and 4, after SLIC processing, an input image is divided into image blocks with close image information, and then, the division information of the first image at the top of the pyramid is expanded to each layer of the pyramid by using an interpolation algorithm, so that all images of the pyramid acquire corresponding image area distribution.
S4, differentiating and distinguishing each subregion: calculating the size, the direction and the color information of the gradient of each pixel of each subregion, performing difference comparison on the 3 pieces of dimensional information and corresponding pixels, and adding 1 to the difference of a point when the difference value of a certain dimension is greater than a preset threshold value; when the information difference values of 3 dimensions are all larger than the threshold value, the difference degree of the point is additionally added with 1, namely the difference value of each pixel point is between 0 and 4.
The image pyramid in the step S2 includes two parameters, namely, a layer height and a layer number, the layer number determines the resolution of the images in the pyramid, the layer height determines the number of the images after filtering processing under a single resolution, and the shape of the pyramid is adjusted through the two parameters, so that the ratio of the detection area at the difference to the total detection area in the subsequent processing is the largest and is close to 100% as much as possible.
S4 comprises the following steps:
s41, calculating the gradient amplitude and direction of the pixel by using a sobel operator:
I x ＝G x I (1)
I y ＝G y I (2)
wherein I is an input image, G x And G y Sobel operators in x-and y-directions, respectively, I x And I y As a gradient map in the corresponding direction, I G Is the magnitude of the gradient, I θ Is the direction of the gradient;
s42, converting the image into an HSV color space, and acquiring the value of the H dimension, namely the color information I of the pixel H Respectively obtaining I of source image pixel and target image pixel G 、I θ And I H After 3 dimensions of information, calculating difference values corresponding to 3 dimensions:
wherein, I' G 、I′ θ 、I′ H Is the corresponding 3-dimensional information value of the source image,
corresponding to the target image 3Value of dimension information, when S G 、S θ 、S H Respectively greater than a gradient threshold tau G 、τ θ 、τ H The total difference D of the pixels p Respectively adding 1, otherwise, when S G 、S θ 、S H When all three are greater than the corresponding threshold value, D p Then additionally adding 1;
s43, summing and normalizing the difference degrees of all pixels in the region to obtain the total difference degree D of the region t ：
Wherein n is the number of regional pixels, D pi Is the ith pixel D p Value of the total difference D of the areas t Greater than a threshold τ t If so, the region is marked as 1, namely a difference region possibly exists, otherwise, the table is marked as 0; after marking all image areas in the pyramid, when the number of images marked as 1 in a certain area is larger than the total number of images tau r If so, determining that the region has a difference;
in formula (9), M k The label value for the region that is indexed by k for the final disparity map, m is the total number of pyramid images,
the marker value of the k region of the ith image.
Referring to fig. 5, it can be seen that the pyramid image has the effect after the different regions are merged, and because the image blocks are obtained by clustering rather than simply dividing, the boundaries of the different regions are more fit and accurate, so that the difference mask is more accurate.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (4)
1. The image difference region detection method based on the super-pixel segmentation is characterized by comprising the following steps of:
s1, registering images to be compared by using a sift algorithm, changing the images to be compared to the same plane, and corresponding coordinates of corresponding points one by one;
s2, constructing respective image pyramids of the source image and the target image through a Gaussian function;
s3, extracting a first image at the topmost layer of the target image pyramid, performing SLIC processing on the first image, and dividing the image into a plurality of sub-regions with relatively consistent information;
s4, differentiating and distinguishing each subregion: calculating the size, the direction and the color information of the gradient of the pixel of each subarea, performing difference comparison on the 3 pieces of dimension information and the corresponding pixels, and adding 1 to the difference of the point when the difference value of a certain dimension is greater than a preset threshold value; when the information difference values of 3 dimensions are all larger than the threshold value, the difference degree of the point is additionally added with 1, namely the difference value of each pixel point is between 0 and 4.
2. The method according to claim 1, wherein the image pyramid in S2 includes two parameters, i.e. a layer height and a layer number, the layer number determines the resolution of how many images are in the pyramid, the layer height determines the number of the filtered images at a single resolution, and the shape of the pyramid is adjusted by the two parameters, so that the ratio of the detection area at the difference to the total detection area in the subsequent processing is the largest.
3. The method according to claim 2, further comprising, after S3, using an interpolation algorithm to expand the segmentation information of the first image at the top of the pyramid onto each layer of the pyramid, so that all images of the pyramid obtain a corresponding and consistent image region distribution.
4. The method for detecting the image difference region based on the super-pixel segmentation as claimed in claim 1, wherein the step S4 comprises the steps of:
s41, calculating the gradient amplitude and direction of the pixel by using a sobel operator:
I x ＝G x I (1)
I y ＝G y I (2)
wherein I is an input image, G x And G y Sobel operators in the x-direction and y-direction, I x And I y As a gradient map in the corresponding direction, I G Is the magnitude of the gradient, I θ Is the direction of the gradient;
s42, converting the image into an HSV color space, and acquiring the value of the H dimension, namely the color information I of the pixel H Respectively obtaining I of source image pixel and target image pixel G 、I θ And I H After 3 dimensions of information, calculating difference values corresponding to 3 dimensions:
wherein, I' G 、I′ θ 、I′ H For the corresponding 3-dimensional information value of the source image,
is the 3-dimensional information value corresponding to the target image, when S G 、S θ 、S H Respectively greater than a gradient threshold tau G 、τ θ 、τ H The total difference D of the pixels p Respectively adding 1, otherwise, when S is not added G 、S θ 、S H When all three are greater than the corresponding threshold value, D p Then additionally adding 1;
s43, summing and normalizing the difference degrees of all pixels in the region to obtain the total difference degree D of the region t ：
Wherein n is the number of regional pixels, D pi Is the ith pixel D p Value of, when the area is totally different by degree D t Greater than a threshold τ t If so, marking the area as 1, namely, a difference area possibly exists, otherwise, recording the table as 0; after marking all image areas in the pyramid, when the number of images marked as 1 in a certain area is larger than the total number of images tau r If so, determining that the region has a difference;
in formula (9), M k The label value for the region that is indexed by k for the final disparity map, m is the total number of pyramid images,
the marker value of the k region of the ith image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211432673.5A CN115829943A (en) | 2022-11-16 | 2022-11-16 | Image difference region detection method based on super-pixel segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211432673.5A CN115829943A (en) | 2022-11-16 | 2022-11-16 | Image difference region detection method based on super-pixel segmentation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115829943A true CN115829943A (en) | 2023-03-21 |
Family
ID=85528366
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211432673.5A Pending CN115829943A (en) | 2022-11-16 | 2022-11-16 | Image difference region detection method based on super-pixel segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115829943A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117372437A (en) * | 2023-12-08 | 2024-01-09 | 安徽农业大学 | Intelligent detection and quantification method and system for facial paralysis |
-
2022
- 2022-11-16 CN CN202211432673.5A patent/CN115829943A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117372437A (en) * | 2023-12-08 | 2024-01-09 | 安徽农业大学 | Intelligent detection and quantification method and system for facial paralysis |
| CN117372437B (en) * | 2023-12-08 | 2024-02-23 | 安徽农业大学 | Intelligent detection and quantification method and system for facial paralysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111243032B (en) | Full-automatic detection method for checkerboard corner points | |
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| CN103325112B (en) | Moving target method for quick in dynamic scene | |
| CN109859226B (en) | Detection method of checkerboard corner sub-pixels for graph segmentation | |
| CN109271937B (en) | Sports ground marker identification method and system based on image processing | |
| CN106683119B (en) | Moving vehicle detection method based on aerial video image | |
| CN108986152B (en) | Foreign matter detection method and device based on difference image | |
| CN106446894A (en) | Method for recognizing position of spherical object based on contour | |
| CN111444778A (en) | Lane line detection method | |
| CN112364865B (en) | Method for detecting small moving target in complex scene | |
| CN104574401A (en) | Image registration method based on parallel line matching | |
| CN109472770B (en) | Method for quickly matching image characteristic points in printed circuit board detection | |
| CN114022439B (en) | Flexible circuit board defect detection method based on morphological image processing | |
| CN104123554A (en) | SIFT image characteristic extraction method based on MMTD | |
| CN112183325B (en) | Road vehicle detection method based on image comparison | |
| CN111695373A (en) | Zebra crossing positioning method, system, medium and device | |
| CN115829943A (en) | Image difference region detection method based on super-pixel segmentation | |
| CN108038458B (en) | Method for automatically acquiring outdoor scene text in video based on characteristic abstract diagram | |
| CN111290582B (en) | Projection interaction area positioning method based on improved linear detection | |
| CN113409334A (en) | Centroid-based structured light angle point detection method | |
| Pan et al. | An efficient method for skew correction of license plate | |
| CN106446832B (en) | Video-based pedestrian real-time detection method | |
| CN117496401A (en) | Full-automatic identification and tracking method for oval target points of video measurement image sequences | |
| CN115311293B (en) | Rapid matching method for printed matter pattern | |
| CN116863458A (en) | License plate recognition method, device, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |