CN109753957A - Image significance detection method, device, storage medium and electronic equipment - Google Patents
Image significance detection method, device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN109753957A CN109753957A CN201811497718.0A CN201811497718A CN109753957A CN 109753957 A CN109753957 A CN 109753957A CN 201811497718 A CN201811497718 A CN 201811497718A CN 109753957 A CN109753957 A CN 109753957A
- Authority
- CN
- China
- Prior art keywords
- subregion
- image
- determined
- boolean
- saliency value
- 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.)
- Granted
Links
Abstract
This disclosure relates to which a kind of image significance detection method, device, storage medium and electronic equipment obtain more accurate saliency value calculated result to reduce the calculation amount in image saliency value calculating process.The image significance detection method comprises determining that ROI region interested and multiple subregions relevant to the ROI region on target image；Determine color characteristic, spatial relation characteristics and the area of each subregion；According to the color characteristic and the area, the contrast metric of each subregion is determined；According to the color characteristic and the spatial relation characteristics, the provincial characteristics of each subregion is determined；Based on the contrast metric and provincial characteristics of each subregion, the saliency value of the target image is determined.
Description
Technical field
This disclosure relates to technical field of image processing, and in particular, to a kind of image significance detection method, is deposited device
Storage media and electronic equipment.
Background technique
The vision system of the mankind can filter out unnecessary background letter when facing complex scene in a very short period of time
Breath, rapidly and accurately obtains target.Therefore, this vision noticing mechanism is used for reference, in technical field of image processing, image occurs
Conspicuousness detection algorithm, so as to the quick obtaining important information in the image data of magnanimity.
But conspicuousness detection algorithm in the related technology is usually to be directed to each of image pixel to carry out significantly
Value calculates, and calculation amount is larger.Also, if the scene in image is more complex, according to conspicuousness detection algorithm in the related technology,
It is unable to get an accurate conspicuousness testing result.
Summary of the invention
Purpose of this disclosure is to provide a kind of image significance detection method, device, storage medium and electronic equipments, with solution
The problem that certainly calculation amount is larger existing for conspicuousness detection algorithm in the related technology and result is inaccurate.
To achieve the goals above, in a first aspect, the disclosure provides a kind of image significance detection method, comprising:
Determine the ROI region interested on target image and multiple subregions relevant to the ROI region, wherein
Subregion relevant to the ROI region, which refers on the target image, at least has part same pixel with the region ROI
The subregion of point；
Determine color characteristic, spatial relation characteristics and the area of each subregion；
According to the color characteristic and the area, the contrast metric of each subregion is determined；
According to the color characteristic and the spatial relation characteristics, the provincial characteristics of each subregion is determined；
Based on the contrast metric and provincial characteristics of each subregion, the saliency value of the target image is determined.
Optionally, described according to the color characteristic and the spatial relation characteristics, determine the area of each subregion
Characteristic of field, comprising:
According to the color characteristic, the color distance between each subregion is determined；
According to the spatial relation characteristics, the space length between each subregion is determined；
According to the color distance and the space length, the provincial characteristics of each subregion is determined.
Optionally, it according to the color distance and the space length, determines the provincial characteristics of each subregion, wraps
It includes:
The region of each subregion is determined according to the color distance and the space length according to following formula
Feature:
Wherein, SdFor the provincial characteristics of subregion, n is the number of multiple subregions relevant to the ROI region, dCol
(Ri,Rj) it is subregion RiWith subregion RjBetween color distance, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space away from
From,To preset weight parameter.
Optionally, the target image is to be obtained according to the size relation of pixel each in image and presetted pixel threshold value
Boolean's image the contrast metric of each subregion is determined according to the color characteristic and the area, comprising:
According to following formula, according to the color characteristic and the area, determine that the contrast of each subregion is special
Sign:
Wherein,For the color characteristic of subregion, θ is the presetted pixel threshold value, and A is the area of subregion.
Optionally, boolean's image is multilayer boolean's image under single image channel, is based on each sub-district
The contrast metric and provincial characteristics in domain, determine the saliency value of the target image, comprising:
It is determined according to following formula based on the contrast and provincial characteristics of each subregion in every layer of boolean's image
The saliency value of the target image:
Wherein, S (x, y) is the saliency value of the target image, and m is the number of plies of boolean's image,For kth layer Boolean Graphs
The default weight parameter of picture, Sk(x, y) is respectively by the contrast metric of the subregion each in every layer of boolean's image and area
The result that characteristic of field is summed after being multiplied.
Optionally, boolean's image is single layer boolean's image under multiple images channel, is based on each sub-district
The contrast metric and provincial characteristics in domain, determine the saliency value of the target image, comprising:
Based on the contrast metric and regional characteristic value of each subregion, determined under each image channel respectively
Single layer boolean's image saliency value；
Maximum value in multiple saliency value of the determination is determined as to the saliency value of the target image.
Second aspect, the disclosure also provide a kind of saliency detection device, comprising:
Area determination module, for determining ROI region interested on target image and related to the region ROI
Multiple subregions, wherein subregion relevant to the ROI region refers on the target image with the ROI region extremely
Less with the subregion of part same pixel point；
Fisrt feature determining module, for determining color characteristic, spatial relation characteristics and the area of each subregion；
Second feature determining module, for determining each subregion according to the color characteristic and the area
Contrast metric；
Third feature determining module, for determining each described according to the color characteristic and the spatial relation characteristics
The provincial characteristics of subregion；
Saliency value determining module, for contrast metric and provincial characteristics based on each subregion, determine described in
The saliency value of target image.
Optionally, the third feature determining module is used for:
According to the color characteristic, the color distance between each subregion is determined；
According to the spatial relation characteristics, the space length between each subregion is determined；
According to the color distance and the space length, the provincial characteristics of each subregion is determined.
Optionally, the third feature determining module is used for:
The region of each subregion is determined according to the color distance and the space length according to following formula
Feature:
Wherein, SdFor the provincial characteristics of subregion, n is the number of multiple subregions relevant to the ROI region, dCol
(Ri,Rj) it is subregion RiWith subregion RjBetween color distance, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space away from
From,To preset weight parameter.
Optionally, the target image is to be obtained according to the size relation of pixel each in image and presetted pixel threshold value
Boolean's image, the second feature determining module is used for:
According to following formula, according to the color characteristic and the area, determine that the contrast of each subregion is special
Sign:
Wherein,For the color characteristic of subregion, θ is the presetted pixel threshold value, and A is the area of subregion.
Optionally, boolean's image is multilayer boolean's image under single image channel, and the saliency value determines mould
Block is used for:
It is determined according to following formula based on the contrast and provincial characteristics of each subregion in every layer of boolean's image
The saliency value of the target image:
Wherein, S (x, y) is the saliency value of the target image, and m is the number of plies of boolean's image,For kth layer Boolean Graphs
The default weight parameter of picture, Sk(x, y) is respectively by the contrast metric of the subregion each in every layer of boolean's image and area
The result that characteristic of field is summed after being multiplied.
Optionally, boolean's image is single layer boolean's image under multiple images channel, and the saliency value determines mould
Block is used for:
Based on the contrast metric and regional characteristic value of each subregion, determined under each image channel respectively
Single layer boolean's image saliency value；
Maximum value in multiple saliency value of the determination is determined as to the saliency value of the target image.
The third aspect, the disclosure also provide a kind of computer readable storage medium, are stored thereon with computer program, the journey
The step of any one the method in first aspect is realized when sequence is executed by processor.
Fourth aspect, the disclosure also provide a kind of electronic equipment, comprising:
Memory is stored thereon with computer program；
Processor, for executing the computer program in the memory, to realize any one institute in first aspect
The step of stating method.
Through the above technical solutions, using each subregion relevant to the ROI region in image as computing unit, rather than
Each of image pixel, to reduce the calculation amount in saliency value calculating process.Also, due to determining saliency value
When combine the spatial relation characteristics of subregion, therefore the image more complex for scene can obtain more accurate saliency value
Calculated result.
Other feature and advantage of the disclosure will the following detailed description will be given in the detailed implementation section.
Detailed description of the invention
Attached drawing is and to constitute part of specification for providing further understanding of the disclosure, with following tool
Body embodiment is used to explain the disclosure together, but does not constitute the limitation to the disclosure.In the accompanying drawings:
Fig. 1 is the flow chart according to a kind of image significance detection method shown in one exemplary embodiment of the disclosure；
Fig. 2 is the process schematic according to the determination image ROI region shown in one exemplary embodiment of the disclosure；
Fig. 3 is a kind of flow chart of image significance detection method shown according to disclosure another exemplary embodiment；
Fig. 4 is according to the image significance detection method and image in the related technology in one exemplary embodiment of the disclosure
The comparative result figure that conspicuousness detection method is respectively handled same image；
Fig. 5 is a kind of block diagram of saliency detection device shown according to one exemplary embodiment of the disclosure；
Fig. 6 is the block diagram according to a kind of electronic equipment shown in one exemplary embodiment of the disclosure.
Specific embodiment
It is described in detail below in conjunction with specific embodiment of the attached drawing to the disclosure.It should be understood that this place is retouched
The specific embodiment stated is only used for describing and explaining the disclosure, is not limited to the disclosure.
The vision system of the mankind can filter out unnecessary background letter when facing complex scene in a very short period of time
Breath, rapidly and accurately obtains target.Therefore, this vision noticing mechanism is used for reference, in technical field of image processing, image occurs
Conspicuousness detection algorithm, so as to the quick obtaining important information in the image data of magnanimity.Saliency detection algorithm
Main process be first determining image saliency value, then the saliency value is added on original image, it is each to amplify on original image
Otherness between a part, to obtain the important information in image.
In the related technology, determine that the saliency value of image carries out saliency value meter primarily directed to each of image pixel
It calculates, calculation amount is larger.It, can not according to conspicuousness detection algorithm in the related technology also, if the scene in image is more complex
Obtain an accurate conspicuousness testing result.
In order to solve the problems in the relevant technologies, the present disclosure proposes a kind of image significance detection method, device, storages
Medium and electronic equipment to reduce the calculation amount in saliency value calculating process, and improve the accurate of saliency testing result
Property.
Fig. 1 is the flow chart according to a kind of image significance detection method shown in one exemplary embodiment of the disclosure.Ginseng
According to Fig. 1, which be may comprise steps of:
Step S101 determines ROI region interested and multiple sub-districts relevant to the ROI region on target image
Domain.
Wherein, subregion relevant to the ROI region refers on the target image and at least has with the region ROI
There is the subregion of part same pixel point.
Step S102 determines color characteristic, spatial relation characteristics and the area of each subregion.
Step S103 determines the contrast metric of each subregion according to the color characteristic and the area.
Step S104 determines the region of each subregion according to the color characteristic and the spatial relation characteristics
Feature.
It should be noted that in the embodiment of the present disclosure for determining the contrast metric and provincial characteristics of each subregion
Sequencing is not construed as limiting.For example, referring to Fig.1, determining every height again after can be the contrast metric for first determining each region
The provincial characteristics in region.Certainly, in other possible modes, it is also possible to after the provincial characteristics for first determining each region again really
Determine the contrast metric of each subregion, or determines the provincial characteristics and contrast metric in each region simultaneously.
Step S105 determines the target image based on the contrast metric and provincial characteristics of each subregion
Saliency value.
The image significance detection method of the embodiment of the present disclosure is to calculate list with each subregion relevant to ROI region
Member, compared in the related technology by each of image pixel be carry out saliency value calculating in a manner of, the embodiment of the present disclosure
Method can reduce the calculation amount in saliency value calculating process.Also, due to combining subregion when determining saliency value
Spatial relation characteristics, therefore the image more complex for scene can obtain more accurate saliency value calculated result.
For the technical solution for making those skilled in the art more understand that the embodiment of the present disclosure provides, below to above-mentioned steps
Carry out illustrated in greater detail.
Illustratively, in step s101, target image can be untreated original image, be also possible to original image
Corresponding channel image, for example, the corresponding LAB channel image of original image or the corresponding RGB channel image of original image,
Etc., the embodiment of the present disclosure is not construed as limiting this.
It should be understood that being convenient for salient region since the feature distribution of each channel image in LAB channel image is uniform
Segmentation, therefore implement the disclosure image significance detection method when, can preferentially select the channel LAB of original image
Image carries out the calculating of saliency value as target image.
By taking target image is the L channel image in LAB triple channel image as an example, it can determine in L channel image first
ROI region.In a kind of possible mode, it can be and first determine the corresponding boolean's image of the L channel image, i.e. bianry image.
Specific method of determination can be first one pixel threshold of setting, and pixel in L channel image is greater than to the pixel of the pixel threshold
The pixel of point is set to 1, and the pixel that pixel in L channel image is less than or equal to the pixel of the pixel threshold is set to 0, thus
The corresponding bianry image of available L channel image.
After obtaining the corresponding bianry image of L channel image, the extraction of connected region can be carried out to the bianry image, is obtained
The image sequence of all connected regions into bianry image.Wherein, connected region refers in bianry image and is connected with image boundary
Connected region.In the embodiments of the present disclosure, the connected region being connected in bianry image with image boundary can be considered figure
As in background, be not conspicuousness detection it needs to be determined that region.Therefore, connected region can be extracted, to reduce
Influence to conspicuousness testing result.For example, connected region extraction, etc., the disclosure can be carried out by zone marker algorithm
Embodiment is not construed as limiting this.Wherein, zone marker algorithm with it is similar in the related technology, which is not described herein again.
In obtaining bianry image after the image sequence of all connected regions, can by image sequence every image with
L channel image is overlapped, and may finally determine the ROI region of L channel image.For example, referring to Fig. 2, scheming B is target image,
Scheming A is the corresponding bianry image of figure B, and has carried out connected region extraction to figure A, has obtained the figure of all connected regions in figure A
As sequence C, then each image schemed in A and image sequence C can be overlapped, figure D be obtained, so as to according to figure
D determines the ROI region of figure B.
It should be understood that in the embodiments of the present disclosure, it, can also be right after obtaining the corresponding bianry image of target image
The bianry image carries out Morphological scale-space, then carries out connected region extraction to the bianry image after Morphological scale-space, with removal
Noise spot in the bianry image simultaneously makes the edge of connected region more smooth, to make the area ROI of determining target image
Domain is more accurate.
After determining the ROI region of target image, every height in multiple subregions relevant to ROI region can be determined
Color characteristic, spatial relation characteristics and the area in region.
Illustratively, subregion can be since the pixel in the upper left corner of target image, as unit of presetted pixel point
It is divided.For example, the length and width of target image is respectively 25 pixels, presetted pixel point is 5, then from
The upper left corner of target image starts, and is unit according to 5 pixels, target image can be divided into 5 × 5 sub-regions.
In the embodiments of the present disclosure, since ROI region is the region bigger than subregion, sub-district relevant to ROI
Domain can have multiple.Also, the subregion having in multiple subregion may be entirely included in ROI region, i.e. the sub-district
Certain of domain and ROI region a part have all identical pixels, and subregion also may only some be included in
In ROI region, i.e., the subregion and the region ROI have part same pixel point.
Therefore, in a kind of possible mode, the area of subregion can be determined according to following formula:
A=∑ f (x, y) (2)
Wherein, A indicates that the area of subregion, (x, y) indicate the coordinate value of each pixel in subregion, RsIndicate ROI
Region.
That is, when pixel is located in ROI region, f (x, y) is 1 by above-mentioned formula (1) and formula (2), and
When pixel is not in ROI region, f (x, y) is 0.Therefore, it can successively calculate every in subregion relevant to ROI region
The f (x, y) of a pixel finally sums to obtained f (x, y), to obtain the area of subregion relevant to ROI region.
In a kind of possible mode, the color characteristic of subregion be can be by calculating all pixels point in subregion
What pixel average determined, it can according to following formula, determine the color characteristic of subregion:
Wherein,Indicate that the color characteristic of subregion, n indicate the pixel number in subregion, fiIt indicates in subregion
The pixel value of ith pixel point.
In a kind of possible mode, the spatial relation characteristics of subregion can be by the centroid of the subregion, i.e. the son
The center point coordinate in region is indicated.Specifically, can determine the spatial relation characteristics (center of subregion according to following formula
Point coordinate):
Wherein, xsIndicate the abscissa of subregion central point, ysIndicate that the ordinate of subregion central point, A indicate sub-district
The area in domain can determine that f (x, y) can determine that x and y respectively indicate sub-district by formula (1) by formula (1) and (2)
The abscissa and ordinate of pixel in domain.
In the manner described above, color characteristic, spatial relation characteristics and the area of subregion relevant to ROI region are determined
Afterwards, the contrast metric of each subregion can be determined, and according to color characteristic and spatial relationship according to color characteristic and area
Feature determines the provincial characteristics of each subregion.
In a kind of possible mode, according to color characteristic and spatial relation characteristics, determine that the region of each subregion is special
Sign can be first according to color characteristic, determine the color distance between each subregion, and according to spatial relation characteristics, determine each
Space length between subregion determines the provincial characteristics of each subregion then according to color distance and space length.
For example, can determine the color distance between each subregion according to following formula according to color characteristic:
Wherein, dCol(Ri,Rj) indicate subregion RiWith subregion RjBetween color distance,Indicate subregion RiColor
Feature,Indicate subregion RjColor characteristic.
For example, can determine the space length between each subregion according to following formula according to spatial relation characteristics:
Wherein, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space length,WithRespectively indicate subregion
RiSpatial relation characteristics (center point coordinate),WithRespectively indicate subregion RjSpatial relation characteristics (central point sit
Mark).
It, can be according to following after determining the color distance and space length between subregion in a kind of possible mode
Formula determines the provincial characteristics of each subregion according to color distance and the space length:
Wherein, SdFor the provincial characteristics of subregion, n is the number of multiple subregions relevant to the ROI region, dCol
(Ri,Rj) it is subregion RiWith subregion RjBetween color distance, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space away from
From,To preset weight parameter.
Illustratively, weight parameter is presetIt can be and be set according to actual conditions, for example, can be by default weight parameterIt is set as 3, etc., the embodiment of the present disclosure is not construed as limiting this.It should be understood that default weight parameterBe set to it is bigger,
Then show that influence of the space length for provincial characteristics is bigger, finally obtained regional characteristic value is smaller.On the contrary, default weight ginseng
NumberIt is set to smaller, then shows that influence of the space length for provincial characteristics is smaller, finally obtained regional characteristic value is bigger.
In a kind of possible mode, when target image is according to the big of pixel each in image and presetted pixel threshold value
When boolean's image of small Relation acquisition, step S103 be can be according to following formula, according to color characteristic and area, be determined each
The contrast metric of subregion:
Wherein,For the color characteristic of subregion, θ is presetted pixel threshold value, and A is the area of subregion.
That is, when target image is bianry image (the boolean's image) obtained in the manner described above, it can basis
The contrast metric of formula (8) calculating each subregion.
It, can be according to the contrast metric and region after the contrast metric and provincial characteristics for determining each subregion
Feature determines the saliency value of target image.
In a kind of possible mode, target can be determined according to following formula according to contrast metric and provincial characteristics
The saliency value of image:
Wherein, S (x, y) indicates the saliency value of target image, and n indicates subregion number relevant to ROI region,Indicate the contrast metric of i-th of subregion,Indicate the provincial characteristics of i-th of subregion.
In alternatively possible mode, target image is multilayer boolean's image under single image channel, for example, right
In L channel image, when setting multiple pixel thresholds, therefore determining boolean's image according to multiple pixel threshold, available L
Multilayer boolean's image under channel.It in this case, can be according to following formula, based on each sub-district in every layer of boolean's image
The contrast and provincial characteristics in domain, determine the saliency value of target image:
Wherein, S (x, y) is the saliency value of target image, and m is the number of plies of boolean's image,For kth layer boolean's image
Default weight parameter, Sk(x, y) is the saliency value of kth layer boolean's image.
Illustratively, default weight parameter can be determined according to following formula
Wherein, NRIndicate the number of connected region in boolean's image.
It should be understood that when target image is multilayer boolean's image under single image channel, it can be for every
Layer boolean's image executes step S101 to step S105 respectively, the saliency value of every layer of boolean's image is obtained, then according to formula
(10) saliency value of the corresponding image in single image channel can be determined.
In alternatively possible mode, target image is single layer boolean's image under multiple images channel, such
In the case of, can first contrast metric and regional characteristic value based on each subregion, determined under each image channel respectively
Then maximum value in obtained multiple saliency value is determined as the saliency value of target image by the saliency value of single layer boolean's image.
Illustratively, target image is single layer boolean's image under LAB triple channel, has been determined respectively first, in accordance with aforesaid way
The saliency value of single layer boolean's image under tri- image channels of LAB, for example, the saliency value of single layer boolean's image under the channel L
0.5, the saliency value of single layer boolean's image under A channel is 0.6, and the saliency value of single layer boolean's image under channel B is 0.7, so
Maximum value in these three saliency value is determined as to the saliency value of target image afterwards, it can be determined as target image for 0.7
Saliency value.
It should be understood that if target image is boolean's image under multiple images channel, and in each image channel
Under obtained multilayer boolean's image, can be to more under each channel image then after the saliency value for determining every layer of boolean's image
Calculating is normalized in the saliency value of layer boolean's image, for example, can be according to following formula, to the multilayer under each image channel
Calculating is normalized in the saliency value of boolean's image:
Wherein, S'(x, y) indicate the saliency value obtained after single layer boolean image normalization calculates, S (x, y) indicates single layer cloth
The former saliency value of your image, Smax(x, y) indicates the maximum value in the saliency value of multilayer boolean's image under each channel image,
Smin(x, y) indicates the minimum value in the saliency value of multilayer boolean's image under each channel image.
Then, the maximum value in multiple saliency value obtained after normalization being calculated is determined as the saliency value of target image,
That is the saliency value of target image may be expressed as:
S (x, y)=max { S '1(x,y),S′2(x,y),S′3(x,y)} (13)
Wherein, S '1(x,y)、S′2(x, y) and S '3(x, y) respectively indicates normalization under multiple images channel and obtains after calculating
Saliency value.
After obtaining the saliency value of target image according to any of the above-described method, which can be added to target figure
As upper, to obtain the corresponding notable figure of the target image.Wherein, specific stacked system with it is similar in the related technology, here
It repeats no more.
Illustrate the image significance detection method of the disclosure with a complete embodiment below.
Referring to Fig. 3, which be may comprise steps of:
Step S301 obtains the channel image under tri- image channels of LAB of original image.
Step S302 determines the corresponding boolean's image of each channel image according to different presetted pixel threshold values respectively.
Step S303 carries out Morphological scale-space to every layer of obtained boolean's image.
Wherein, the detailed process of Morphological scale-space with it is similar in the related technology, which is not described herein again.
Step S304, the ROI region of the image after determining Morphological scale-space and multiple sub-districts relevant to the ROI region
Domain.
Step S305 determines color characteristic, spatial relation characteristics and the area of each subregion.
Step S306 determines the significant of every layer of boolean's image according to color characteristic, spatial relation characteristics and area features
Value.
Step S307 determines each channel figure according to the saliency value of multilayer boolean's image under same figure channel respectively
The saliency value of picture.
Step S308 determines that the maximum value in channel image saliency value is the saliency value of original image.
Step S309, saliency value is added on original image, obtains the corresponding notable figure of original image.
Wherein, the specific implementation process of above-mentioned steps is described in detail above, and which is not described herein again.
The image significance detection method of the embodiment of the present disclosure is to calculate list with each subregion relevant to ROI region
Member, compared in the related technology by each of image pixel be carry out saliency value calculating in a manner of, reduce saliency value
Calculation amount in calculating process.Also, due to the spatial relation characteristics for combining subregion when determining saliency value, for
The more complex image of scene can obtain more accurate saliency value calculated result.
Fig. 4 is right respectively for the image significance detection method of the disclosure and image significance detection method in the related technology
The comparative result figure that same image is handled.Referring to Fig. 4, first row image is the target image of input, and third is classified as
The result figure obtained after being handled by the image significance detection method of the disclosure target image, other, which are classified as, passes through phase
The result figure that image significance detection method in the technology of pass obtains after handling target image.As it can be seen that compared to correlation
Scheme in technology, the image significance detection method of the disclosure have obtained being more clear accurate notable figure, that is, the disclosure
Image significance detection method improves the accuracy of saliency testing result.
Based on the same inventive concept, referring to Fig. 5, the disclosure also provides a kind of saliency detection device 500, can be with
It comprises the following modules:
Area determination module 501, for determine ROI region interested on target image and with the ROI region phase
Close multiple subregions, wherein subregion relevant to the ROI region refer on the target image with the ROI region
At least with the subregion of part same pixel point；
Fisrt feature determining module 502, for determining color characteristic, spatial relation characteristics and the face of each subregion
Product；
Second feature determining module 503, for determining each subregion according to the color characteristic and the area
Contrast metric；
Third feature determining module 504, for determining each institute according to the color characteristic and the spatial relation characteristics
State the provincial characteristics of subregion；
Saliency value determining module 505 determines institute for contrast metric and provincial characteristics based on each subregion
State the saliency value of target image.
Optionally, the third feature determining module 504 is used for:
According to the color characteristic, the color distance between each subregion is determined；
According to the spatial relation characteristics, the space length between each subregion is determined；
According to the color distance and the space length, the provincial characteristics of each subregion is determined.
Optionally, the third feature determining module 504 is used for:
The region of each subregion is determined according to the color distance and the space length according to following formula
Feature:
Wherein, SdFor the provincial characteristics of subregion, n is the number of multiple subregions relevant to the ROI region, dCol
(Ri,Rj) it is subregion RiWith subregion RjBetween color distance, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space away from
From,To preset weight parameter.
Optionally, the target image is to be obtained according to the size relation of pixel each in image and presetted pixel threshold value
Boolean's image, the second feature determining module 503 is used for:
According to following formula, according to the color characteristic and the area, determine that the contrast of each subregion is special
Sign:
Wherein,For the color characteristic of subregion, θ is the presetted pixel threshold value, and A is the area of subregion.
Optionally, boolean's image is multilayer boolean's image under single image channel, and the saliency value determines mould
Block 505 is used for:
It is determined according to following formula based on the contrast and provincial characteristics of each subregion in every layer of boolean's image
The saliency value of the target image:
Wherein, S (x, y) is the saliency value of the target image, and m is the number of plies of boolean's image,For kth layer Boolean Graphs
The default weight parameter of picture, Sk(x, y) is respectively by the contrast metric of the subregion each in every layer of boolean's image and area
The result that characteristic of field is summed after being multiplied.
Optionally, boolean's image is single layer boolean's image under multiple images channel, and the saliency value determines mould
Block 505 is used for:
Based on the contrast metric and regional characteristic value of each subregion, determined under each image channel respectively
Single layer boolean's image saliency value；
Maximum value in multiple saliency value of the determination is determined as to the saliency value of the target image.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
The saliency detection device of the embodiment of the present disclosure is to calculate list with each subregion relevant to ROI region
Member can reduce the calculation amount in saliency value calculating process.Also, due to the space for combining subregion when determining saliency value
Relationship characteristic, therefore the image more complex for scene can obtain more accurate saliency value calculated result.
Based on the same inventive concept, the disclosure also provides a kind of electronic equipment, comprising:
Memory is stored thereon with computer program；
Processor, for executing the computer program in the memory, the step of to realize any of the above-described method.
In one exemplary embodiment, the block diagram of the electronic equipment can be as shown in Figure 6.Referring to Fig. 6, the electronic equipment
600 may include: processor 601, memory 602.The electronic equipment 600 can also include multimedia component 603, input/defeated
One or more of (I/O) interface 604 and communication component 605 out.
Wherein, processor 601 is used to control the integrated operation of the electronic equipment 600, to complete above-mentioned saliency
All or part of the steps in detection method.Memory 602 is for storing various types of data to support in the electronic equipment
600 operation, these data for example may include any application or method for operating on the electronic equipment 600
Instruction and the relevant data of application program, such as the color characteristics of each sub-regions, spatial relation characteristics, area etc..It should
Memory 602 can be realized by any kind of volatibility or non-volatile memory device or their combination, such as static state
Random access memory (Static Random Access Memory, abbreviation SRAM), electrically erasable programmable read-only memory
(Electrically Erasable Programmable Read-Only Memory, abbreviation EEPROM), it is erasable to compile
Journey read-only memory (Erasable Programmable Read-Only Memory, abbreviation EPROM), may be programmed read-only storage
Device (Programmable Read-Only Memory, abbreviation PROM), and read-only memory (Read-Only Memory, referred to as
ROM), magnetic memory, flash memory, disk or CD.Multimedia component 603 may include screen and audio component.Wherein
Screen for example can be touch screen, the saliency value for the target image being displayed for, and is shown on target image and folds
Add the notable figure, etc. obtained after saliency value.
I/O interface 604 provides interface between processor 601 and other interface modules, other above-mentioned interface modules can be with
It is keyboard, mouse, button etc..These buttons can be virtual push button or entity button.Communication component 605 is set for the electronics
Wired or wireless communication is carried out between standby 600 and other equipment.Wireless communication, such as Wi-Fi, bluetooth, near-field communication (Near
Field Communication, abbreviation NFC), 2G, 3G or 4G or they one or more of combination, therefore it is corresponding
The communication component 605 may include: Wi-Fi module, bluetooth module, NFC module.
In one exemplary embodiment, electronic equipment 600 can be by one or more application specific integrated circuit
(Application Specific Integrated Circuit, abbreviation ASIC), digital signal processor (Digital
Signal Processor, abbreviation DSP), digital signal processing appts (Digital Signal Processing Device,
Abbreviation DSPD), programmable logic device (Programmable Logic Device, abbreviation PLD), field programmable gate array
(Field Programmable Gate Array, abbreviation FPGA), controller, microcontroller, microprocessor or other electronics member
Part is realized, for executing above-mentioned image significance detection method.
In a further exemplary embodiment, a kind of computer readable storage medium including program instruction is additionally provided, it should
The step of above-mentioned image significance detection method is realized when program instruction is executed by processor.For example, this computer-readable is deposited
Storage media can be the above-mentioned memory 602 including program instruction, and above procedure instruction can be by the processor of electronic equipment 600
601 execute to complete above-mentioned image significance detection method.
The preferred embodiment of the disclosure is described in detail in conjunction with attached drawing above, still, the disclosure is not limited to above-mentioned reality
The detail in mode is applied, in the range of the technology design of the disclosure, a variety of letters can be carried out to the technical solution of the disclosure
Monotropic type, these simple variants belong to the protection scope of the disclosure.
It is further to note that specific technical features described in the above specific embodiments, in not lance
In the case where shield, can be combined in any appropriate way, in order to avoid unnecessary repetition, the disclosure to it is various can
No further explanation will be given for the combination of energy.
In addition, any combination can also be carried out between a variety of different embodiments of the disclosure, as long as it is without prejudice to originally
Disclosed thought equally should be considered as disclosure disclosure of that.
Claims (10)
1. a kind of image significance detection method characterized by comprising
Determine the ROI region interested on target image and multiple subregions relevant to the ROI region, wherein with institute
It states the relevant subregion of ROI region and refers to that at least there is on the target image with the ROI region part same pixel point
Subregion；
Determine color characteristic, spatial relation characteristics and the area of each subregion；
According to the color characteristic and the area, the contrast metric of each subregion is determined；
According to the color characteristic and the spatial relation characteristics, the provincial characteristics of each subregion is determined；
Based on the contrast metric and provincial characteristics of each subregion, the saliency value of the target image is determined.
2. the method according to claim 1, wherein described special according to the color characteristic and the spatial relationship
Sign, determines the provincial characteristics of each subregion, comprising:
According to the color characteristic, the color distance between each subregion is determined；
According to the spatial relation characteristics, the space length between each subregion is determined；
According to the color distance and the space length, the provincial characteristics of each subregion is determined.
3. according to the method described in claim 2, it is characterized in that, being determined according to the color distance and the space length
The provincial characteristics of each subregion, comprising:
The provincial characteristics of each subregion is determined according to the color distance and the space length according to following formula:
Wherein, SdFor the provincial characteristics of subregion, n is the number of multiple subregions relevant to the ROI region, dCol(Ri,
Rj) it is subregion RiWith subregion RjBetween color distance, dDis(Ri,Rj) it is subregion RiWith subregion RjBetween space length,To preset weight parameter.
4. method according to claim 1-3, which is characterized in that the target image is according to each in image
Boolean's image that the size relation of pixel and presetted pixel threshold value obtains is determined according to the color characteristic and the area
The contrast metric of each subregion, comprising:
The contrast metric of each subregion is determined according to the color characteristic and the area according to following formula:
Wherein,For the color characteristic of subregion, θ is the presetted pixel threshold value, and A is the area of subregion.
5. according to the method described in claim 4, it is characterized in that, boolean's image is the multilayer under single image channel
Boolean's image determines the saliency value of the target image based on the contrast metric and provincial characteristics of each subregion, packet
It includes:
According to following formula, based on the contrast and provincial characteristics of each subregion in every layer of boolean's image, determine described in
The saliency value of target image:
Wherein, S (x, y) is the saliency value of the target image, and m is the number of plies of boolean's image,For kth layer boolean's image
Default weight parameter, Sk(x, y) is respectively that the contrast metric of the subregion each in every layer of boolean's image and region is special
The result that sign is summed after being multiplied.
6. according to the method described in claim 4, it is characterized in that, boolean's image is the single layer under multiple images channel
Boolean's image determines the saliency value of the target image based on the contrast metric and provincial characteristics of each subregion, packet
It includes:
Based on the contrast metric and regional characteristic value of each subregion, the list under each image channel is determined respectively
The saliency value of layer boolean's image；
Maximum value in multiple saliency value of the determination is determined as to the saliency value of the target image.
7. a kind of saliency detection device characterized by comprising
Area determination module, for determining ROI region interested on target image and relevant to the ROI region multiple
Subregion, wherein subregion relevant to the ROI region refers on the target image at least to be had with the ROI region
The subregion of part same pixel point；
Fisrt feature determining module, for determining color characteristic, spatial relation characteristics and the area of each subregion；
Second feature determining module, for determining the comparison of each subregion according to the color characteristic and the area
Spend feature；
Third feature determining module, for determining each sub-district according to the color characteristic and the spatial relation characteristics
The provincial characteristics in domain；
Saliency value determining module determines the target for contrast metric and provincial characteristics based on each subregion
The saliency value of image.
8. device according to claim 7, which is characterized in that the third feature determining module is used for:
According to the color characteristic, the color distance between each subregion is determined；
According to the spatial relation characteristics, the space length between each subregion is determined；
According to the color distance and the space length, the provincial characteristics of each subregion is determined.
9. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor
The step of any one of claim 1-6 the method is realized when row.
10. a kind of electronic equipment characterized by comprising
Memory is stored thereon with computer program；
Processor, for executing the computer program in the memory, to realize described in any one of claim 1-6
The step of method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811497718.0A CN109753957B (en) | 2018-12-07 | 2018-12-07 | Image significance detection method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811497718.0A CN109753957B (en) | 2018-12-07 | 2018-12-07 | Image significance detection method and device, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109753957A true CN109753957A (en) | 2019-05-14 |
| CN109753957B CN109753957B (en) | 2020-11-27 |
Family
ID=66402645
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811497718.0A Active CN109753957B (en) | 2018-12-07 | 2018-12-07 | Image significance detection method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109753957B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111899841A (en) * | 2020-08-06 | 2020-11-06 | 北京奥康达体育产业股份有限公司 | Human motion biomechanics monitoring and management system |
| CN111950544A (en) * | 2020-06-30 | 2020-11-17 | 杭州依图医疗技术有限公司 | Method and device for determining interest region in pathological image |
| CN113495301A (en) * | 2020-03-20 | 2021-10-12 | 株式会社理光 | Bandage inspection method, bandage inspection device, storage medium, and system |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008136007A3 (en) * | 2007-05-08 | 2010-02-25 | Amihay Halamish | Acquiring regions of interest at a high frame rate |
| EP2273424A1 (en) * | 2009-07-08 | 2011-01-12 | Honeywell International Inc. | Automated target detection and recognition system and method |
| CN103093454A (en) * | 2012-12-20 | 2013-05-08 | 杭州电子科技大学 | Center periphery surrounding optimization method for testing visual significance |
| CN104103082A (en) * | 2014-06-06 | 2014-10-15 | 华南理工大学 | Image saliency detection method based on region description and priori knowledge |
| CN104867130A (en) * | 2015-04-21 | 2015-08-26 | 浙江工业大学 | Self-adaptive segmentation method based on crack image subarea gray scale mean value |
| US9256806B2 (en) * | 2010-03-19 | 2016-02-09 | Digimarc Corporation | Methods and systems for determining image processing operations relevant to particular imagery |
| CN106570498A (en) * | 2016-10-12 | 2017-04-19 | 中国科学院自动化研究所 | Salient region detection method and system |
| CN106611427A (en) * | 2015-10-21 | 2017-05-03 | 中国人民解放军理工大学 | A video saliency detection method based on candidate area merging |
| EP3179442A1 (en) * | 2015-12-09 | 2017-06-14 | Thomson Licensing | Method for tracking a region of interest, computer readable storage medium and apparatus configured to track a region of interest |
| CN107016682A (en) * | 2017-04-11 | 2017-08-04 | 四川大学 | A kind of notable object self-adapting division method of natural image |
| CN108596921A (en) * | 2018-05-10 | 2018-09-28 | 苏州大学 | Method, apparatus, equipment and the readable storage medium storing program for executing of salient region of image detection |
| CN108921820A (en) * | 2018-05-30 | 2018-11-30 | 咸阳师范学院 | A kind of saliency object detection method based on feature clustering and color contrast |
-
2018
- 2018-12-07 CN CN201811497718.0A patent/CN109753957B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008136007A3 (en) * | 2007-05-08 | 2010-02-25 | Amihay Halamish | Acquiring regions of interest at a high frame rate |
| EP2273424A1 (en) * | 2009-07-08 | 2011-01-12 | Honeywell International Inc. | Automated target detection and recognition system and method |
| US9256806B2 (en) * | 2010-03-19 | 2016-02-09 | Digimarc Corporation | Methods and systems for determining image processing operations relevant to particular imagery |
| CN103093454A (en) * | 2012-12-20 | 2013-05-08 | 杭州电子科技大学 | Center periphery surrounding optimization method for testing visual significance |
| CN104103082A (en) * | 2014-06-06 | 2014-10-15 | 华南理工大学 | Image saliency detection method based on region description and priori knowledge |
| CN104867130A (en) * | 2015-04-21 | 2015-08-26 | 浙江工业大学 | Self-adaptive segmentation method based on crack image subarea gray scale mean value |
| CN106611427A (en) * | 2015-10-21 | 2017-05-03 | 中国人民解放军理工大学 | A video saliency detection method based on candidate area merging |
| EP3179442A1 (en) * | 2015-12-09 | 2017-06-14 | Thomson Licensing | Method for tracking a region of interest, computer readable storage medium and apparatus configured to track a region of interest |
| CN106570498A (en) * | 2016-10-12 | 2017-04-19 | 中国科学院自动化研究所 | Salient region detection method and system |
| CN107016682A (en) * | 2017-04-11 | 2017-08-04 | 四川大学 | A kind of notable object self-adapting division method of natural image |
| CN108596921A (en) * | 2018-05-10 | 2018-09-28 | 苏州大学 | Method, apparatus, equipment and the readable storage medium storing program for executing of salient region of image detection |
| CN108921820A (en) * | 2018-05-30 | 2018-11-30 | 咸阳师范学院 | A kind of saliency object detection method based on feature clustering and color contrast |
Non-Patent Citations (3)
| Title |
|---|
| PANDIVALAVAN MANIPOONCHELVI,ET AL: "《Region-based saliency detection》", 《IET IMAGE PROCESS》 * |
| 曾祥鑫，等: "《基于对比度和局部结构特征的显著性检测》", 《重庆理工大学学报（自然科学）》 * |
| 钱堃，等: "《基于颜色和空间距离的显著性区域固定阈值分割算法》", 《计算机科学》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113495301A (en) * | 2020-03-20 | 2021-10-12 | 株式会社理光 | Bandage inspection method, bandage inspection device, storage medium, and system |
| CN111950544A (en) * | 2020-06-30 | 2020-11-17 | 杭州依图医疗技术有限公司 | Method and device for determining interest region in pathological image |
| CN111899841A (en) * | 2020-08-06 | 2020-11-06 | 北京奥康达体育产业股份有限公司 | Human motion biomechanics monitoring and management system |
| CN111899841B (en) * | 2020-08-06 | 2021-05-11 | 北京奥康达体育产业股份有限公司 | Human motion biomechanics monitoring and management system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109753957B (en) | 2020-11-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109753957A (en) | Image significance detection method, device, storage medium and electronic equipment | |
| US11410277B2 (en) | Method and device for blurring image background, storage medium and electronic apparatus | |
| CN109325954B (en) | Image segmentation method and device and electronic equipment | |
| US10410084B2 (en) | Devices, systems, and methods for anomaly detection | |
| EP2927873B1 (en) | Image processing apparatus and image processing method | |
| US9798951B2 (en) | Apparatus, method, and processor for measuring change in distance between a camera and an object | |
| CN109359602A (en) | Method for detecting lane lines and device | |
| US8842908B2 (en) | Method of separating front view and background and apparatus | |
| US9911058B2 (en) | Method, system and apparatus for updating a scene model | |
| CN109241345B (en) | Video positioning method and device based on face recognition | |
| CN111127543B (en) | Image processing method, device, electronic equipment and storage medium | |
| US10885636B2 (en) | Object segmentation apparatus and method using Gaussian mixture model and total variation | |
| CN110648363A (en) | Camera posture determining method and device, storage medium and electronic equipment | |
| CN109035158A (en) | Image fuzzy processing method, device, storage medium and electronic equipment | |
| JP7379684B2 (en) | Image generation method and device and computer program | |
| CN107451976A (en) | A kind of image processing method and device | |
| CN107018390A (en) | The control method of image processing equipment and image processing equipment | |
| WO2014123583A1 (en) | Image processor with edge selection functionality | |
| US9317904B2 (en) | Image processing system and image processing method | |
| CN108960012B (en) | Feature point detection method and device and electronic equipment | |
| US9904981B2 (en) | Image processing apparatus and method for color-depth demosaicing | |
| CN111028169A (en) | Image correction method, device, terminal equipment and storage medium | |
| US9304598B2 (en) | Mobile terminal and method for generating control command using marker attached to finger | |
| WO2022206679A1 (en) | Image processing method and apparatus, computer device and storage medium | |
| US10354142B2 (en) | Method for holographic elements detection in video stream |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |