CN113192092B - Contour detection method for simulating fusion of properties of receptor field of XYW cell - Google Patents
Contour detection method for simulating fusion of properties of receptor field of XYW cell Download PDFInfo
- Publication number
- CN113192092B CN113192092B CN202110493954.0A CN202110493954A CN113192092B CN 113192092 B CN113192092 B CN 113192092B CN 202110493954 A CN202110493954 A CN 202110493954A CN 113192092 B CN113192092 B CN 113192092B
- Authority
- CN
- China
- Prior art keywords
- response
- channel
- contour
- image
- cells
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 30
- 230000004927 fusion Effects 0.000 title claims abstract description 19
- 230000004044 response Effects 0.000 claims abstract description 182
- 230000001629 suppression Effects 0.000 claims abstract description 18
- 230000002401 inhibitory effect Effects 0.000 claims abstract description 14
- 230000002146 bilateral effect Effects 0.000 claims abstract description 7
- 238000001914 filtration Methods 0.000 claims abstract description 7
- 238000000034 method Methods 0.000 claims description 31
- 230000006870 function Effects 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 19
- 230000000007 visual effect Effects 0.000 claims description 18
- 230000005764 inhibitory process Effects 0.000 claims description 16
- 238000004364 calculation method Methods 0.000 claims description 13
- 239000000126 substance Substances 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 claims description 4
- 238000006467 substitution reaction Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000012790 confirmation Methods 0.000 claims description 3
- 230000002964 excitative effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000002093 peripheral effect Effects 0.000 claims description 3
- 230000000638 stimulation Effects 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 2
- 210000004027 cell Anatomy 0.000 description 39
- 230000007246 mechanism Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000010791 quenching Methods 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 239000011664 nicotinic acid Substances 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a contour detection method for simulating the fusion of properties of a receptor field of XYW cells, which comprises the following steps: A. acquiring a gray image and a parallax image of an image to be detected; B. carrying out rapid bilateral filtering on the gray level image of the image to be detected; C. the filtered image is processed by X, Y and W channels respectively; D. inhibiting through sparse coding, and removing redundant textures; E. summing the suppressed X-channel contour response, Y-channel contour response and W-channel contour response with the parallax contour response of the corresponding channel of the parallax image respectively; F. fusing the X channel contour response, the Y channel contour response and the W channel contour response of the strengthened weak contour by using different watching planes to obtain a final contour and an optimal direction; G. and carrying out non-maximum suppression on the final contour response by using the optimal direction to obtain a suppressed final contour. The invention enhances the integrity and continuity of the contour and further improves the detection performance of the contour.
Description
Technical Field
The invention relates to the field of image processing, in particular to a contour detection method for simulating fusion of XYW cell receptive field characteristics.
Background
In real life, the target contour is often embedded in a complex background, and it is a difficult task to distinguish the target contour from the background texture well. The following three neuronal cells with different space summation characteristics exist in the biological visual nerve mechanism: x-type cells, Y-type cells, and W-type cells. However, the existing contour detection model based on biological visual mechanism only considers the action mechanism of X-type cells or Y-type cells singly, neglects the action of W-type cells in contour detection, and does not consider the effective fusion of the response characteristics of the three types of cells. And the lack of important characteristic clues of disparity information causes incomplete and discontinuous contour detection in previous work, thereby reducing the performance of a contour detection model.
Disclosure of Invention
The invention aims to provide a contour detection method for simulating fusion of receptive field characteristics of XYW cells, and provides a bionic contour detection model based on a unique response mechanism of X, Y and W neuron cells to different visual stimuli in a V1 region and an information fusion method thereof.
The technical scheme of the invention is as follows:
the contour detection method for simulating the fusion of the properties of the receptor field of the XYW cells comprises the following steps:
A. acquiring a gray image and a parallax image of an image to be detected;
B. carrying out rapid bilateral filtering on a gray level image of an image to be detected to obtain a filtered image;
C. processing the filtered image by X, Y and W channels respectively to obtain X channel contour response, Y channel contour response and W channel contour response;
D. respectively suppressing the X channel contour response, the Y channel contour response and the W channel contour response through sparse coding, removing redundant textures, and obtaining the suppressed X channel contour response, Y channel contour response and W channel contour response;
E. respectively summing the suppressed X-channel contour response, Y-channel contour response and W-channel contour response with the parallax contour response of the corresponding channel of the parallax image to obtain the X-channel contour response, Y-channel contour response and W-channel contour response with strengthened weak contours;
F. fusing the X channel contour response, the Y channel contour response and the W channel contour response which strengthen the weak contour by using different watching planes to obtain a final contour and an optimal direction;
G. and carrying out non-maximum suppression on the final contour response by using the optimal direction to obtain a suppressed final contour.
In step B, the formula of the fast bilateral filtering is:
f g (x，y)＝exp(μ· - f b (x，y)+f d (x，y)) (1)
f d (x，y)＝log(I(x，y)-f b (x，y)) (3)
where I (x, y) represents a gray image, σ represents a standard deviation of a Gaussian function, and f b (x, y) represents an image of the gray-scale image after smoothing by a Gaussian function, f d (x, y) is represented by a gray image and f b (x, y) is a difference and logarithmic to obtain an image, and μ represents f b (x, y) and f d The (x, y) link strength, symbol · represents multiplication, and x represents convolution.
In the step C, the formula for obtaining the X channel contour response is:
FR X⌒C (x，y)＝E X⌒C (x，y)-α X⌒C ·T X⌒C (x，y) (4)
wherein: e X⌒C (X, y) is the inhibitory response of the classical receptor field of the X channel, T X⌒C (X, y) is the inhibitory response of the X-channel nonclassical receptive field, α X⌒C The inhibition intensity is shown.
The inhibition response T for obtaining the X channel classical receptive field X⌒C The process of (x, y) is as follows:
a. the classical receptive field of X cells was simulated using the first derivative of gaussian with directional selectivity:
wherein, the rotation angle theta represents the direction parameter 0 epsilon [0, pi) of the receptive field, and gamma represents the constant gamma =0.5 of the ratio of the long axis and the short axis of the receptive field;
b. order to
Obtaining the energy response E of the classical receptive field of the X cells to the external stimulus X⌒C (x，y；θ i )：
E X⌒C (x，y；θ i )＝f C (x，y)*RF(x，y；θ i ，σ XC ) (8)
Wherein C belongs to { g, d }, g represents a gray level image, and d represents a parallax image;
theta as above i Each pixel point is represented as N θ The direction angle in each direction is calculated as follows:
c. comparing the energy information E of each pixel point in different direction angles X⌒C (x，y；θ i ) (ii) a Obtaining the optimal response E of the current position X⌒C (x, y) obtaining an inhibitory response of the classical receptive field; simultaneously obtaining the corresponding optimal direction angle theta X⌒g (x，g)：
E X⌒C (x，y)＝max{E X⌒C (x，y；θ i )|i＝1，2，…，N θ } (10)
The inhibition response E for acquiring the X channel non-classical receptive field X⌒C The process of (x, y) is as follows:
a. optimal energy response and suppression weight omega of each pixel point in classical receptive field u (x, y) convolution to obtain the inhibitory response T of the non-classical receptive field X⌒C (x，y)：
T X⌒C (x，y)＝E X⌒C (x，y)*ω u (x，y) (12)
Wherein | | xi | purple 1 Represents L 1 Norm, H (Z) = max (0, Z);
DOG is a two-dimensional Gaussian difference function, and the formula is as follows:
wherein σ u Standard deviation of the non-classical receptive field is indicated and k represents the scale ratio of the peripheral inhibitory region to the central excitatory region.
In the step C, the formula for obtaining the Y channel contour response is:
FR Y⌒C (x，y)＝S Y⌒C (x，y)-α Y⌒C ·T Y⌒C (x，y) (15)
wherein S is Y⌒C (x, Y) is the Y-channel nonlinear quench response, T Y⌒C (x, Y) is the Y-channel linear suppression response, α Y⌒C The inhibition intensity is shown.
Obtaining Y-channel nonlinear suppression response S Y⌒C The process of (x, y) is as follows:
a. by sigma Y Substitution of σ in equations 1-3, 5, 6 X Obtaining the classical receptive field response E of the Y cell Y⌒C (x，y；θ i ) Then E is added Y⌒C (x，y；θ i ) Substituting into formula 9-11 to obtain the optimal response E of Y cell Y⌒C (x, y) and corresponding optimal direction angle θ y⌒g (x，y)；
b. Setting a region S of finite size xy Each pixel point in the region corresponds to the central position of one subunit, the rectification property of the subunit is simulated by a Gaussian function, and then, each pixel point is subjected to rectificationSumming the contributions of the subunits results in a non-linear subunit response S Y⌒C (x, y) in the formula:
wherein the content of the first and second substances,
representing a window S centered on (x, y) xy Intrinsic response E Y_C Information in (x, y). Sigma s ＝σ YC Calculating the spatial standard deviation of the subunit size; s xy Radius of (d) size r s ＝12·σ YC (ii) a N is
The number of pixels of (a); v represents the ordinate of the Y-type cell subunit;
the acquisition of the linear suppressive response T of the Y channel Y⌒C The process of (x, y) is as follows: will S Y⌒C (x, y) instead of E X⌒C (x, y) is calculated by substituting (x, y) as an input into equation 12.
In the step C, the calculation process for obtaining the W channel contour response is as follows:
let σ = σ n - Will σ n Substituting into formula 5 to obtain Gaussian first derivative function RF (x, y; theta, sigma) with different scales and shapes n ) (ii) a Standard deviation sigma n The calculation formula of (a) is as follows:
σ WC is the standard deviation of CRF for W cells, since CRF is similar for W cells and X cells, here σ WC ＝σ XC (ii) a S represents the classical perception of W cellsThe size of the field scale change; n denotes the number of classical receptive field scale changes, where n =25/σ WC ；
b. From a Gaussian first derivative function RF (x, y; theta, sigma) n ) Selecting different micro-motion areas, wherein the selection formula of the micro-motion areas is as follows:
RF (x, y; theta, sigma) n ) Multiplying by equation 19 yields a gaussian first derivative function with fixation and fixation micromotion characteristics:
RM n (x，y；θ i ，σ n )＝RF(x，y；θ i ，σ n )·M n (x，y) (20)；
c. the response of the classical receptive field under the visual micromotion to the external stimulus is input from the characteristic diagram and RM n (x，y；θ i ，σ n ) Convolution yields an energy response E W⌒C (x，y；θ i ，σ n )：
E W⌒C (x，y；θ i ，σ n )＝f C (x，y)*RM n (x，y；θ i ，σ n ) (21)；
d. From E W⌒C (x，y；θ i ，σ n ) In place of E X⌒C (x，y；θ i ) Substituting into the formula 9-11 to calculate the optimal response E of W cells W⌒C (x, y) and corresponding optimal direction angle θ W⌒g (x，y)；
From E W⌒C (x, y) instead of E X⌒C (x, y) as input into equation 12 to calculate the non-classical receptor suppression response T of W cells W⌒C (x，y)；
e. W cell contour response FR W⌒C The formula for the calculation (x, y) is:
FR W⌒C (x，y)＝E W⌒C (x，y)-α W⌒C ·T W⌒C (22)
wherein E is W⌒C (x, y) is the optimal response of W cells, T X⌒C (x, y) is WtongInhibitory response of the Dow nonclassical receptor field, alpha W⌒C The inhibition intensity is shown.
The sparse coding formula in the step D is as follows:
wherein e is M (x，y)∈{FR M⌒g (X, Y) }, namely M belongs to { X, Y, W } and represents the stimulation response under X, Y and W channels based on the gray level image information;
is a histogram of local gradient amplitude values of each information channel centered at (x, y), and t is
Dimension of (d); i 1 Is L 1 Norm, | | 2 Is L 2 A norm; />
Represents->
Min represents the minimum value of the two;
applying sparse coding to inhibit, and obtaining formulas of inhibited X channel contour response, Y channel contour response and W channel contour response as follows;
LR M⌒g (x，y)＝sm M (x，y）·FR M⌒g (x，y) (24)。
in the step E, the suppressed X-channel contour response, Y-channel contour response and W-channel contour response are respectively compared with the parallax contour response FR of the corresponding channel of the parallax image M⌒d The formula for the sum (x, y) is:
R M ＝FR M⌒d (x，y)+LR M⌒g (x，y) (25)。
in the step F, the process of fusing the X channel contour response, the Y channel contour response and the W channel contour response which strengthen the weak contour by using different gazing planes is as follows:
a. the visual information of each watching plane is processed by an X channel, the visual information of a near plane is processed by a Y channel, the visual information of a far plane is processed by a W channel, and the confirmation formula of the watching planes is as follows:
b. normalizing the parallax image to [0, P]Parallax range of (1), using d (x, y) represents the parallax image after normalization; setting a parallax range on a watching plane as beta; when round (G (x, y), 1) =0.5, it represents a near plane; when round (G (x, y), 1) = -0.5 is set, a far plane is represented; setting G (x, y) to other values, indicating on the gaze plane;
c. and fusing visual information on X, Y and W channels through three gazing planes to obtain a final contour R (X, Y) and an optimal direction theta (X, Y):
the invention designs a unique simulation function, models the respective unique physiological characteristic response of X, Y and W cells, respectively and simultaneously processes the gray image and the parallax image, and finally effectively fuses the contour characteristics of X, Y and W channels according to the fixation surface determined by the parallax image. Experimental results show that the contour detection method based on the model can better preserve the integrity of the contour and inhibit background textures, and has good application prospect.
Drawings
FIG. 1 is a schematic diagram of the fusion process of three channels X, Y and W according to three fixation surfaces in example 1;
FIG. 2 is a graph showing the comparison between the contour detection method of example 1 and the contour detection method of reference 1
Detailed Description
Example 1
1. The contour detection method for simulating the fusion of the properties of the receptor field of the XYW cell provided by the embodiment comprises the following steps:
the contour detection method for simulating the fusion of the properties of the receptor field of the XYW cell comprises the following steps:
A. acquiring a gray image and a parallax image of an image to be detected;
B. carrying out rapid bilateral filtering on the gray level image of the image to be detected to obtain a filtered image; the formula for fast bilateral filtering is:
f g (x，y)＝exp(μ· - f b (x，y)+f d (x，y)) (1)
f d (x，y)＝log(I(x，y)-f b (x，y)) (3)
wherein I (x, y) represents a grayscale image; σ represents a standard deviation of the Gaussian function; f. of b (x, y) represents an image after the gray image is smoothed by a gaussian function; f. of d (x, y) is represented by a gray image and f b (x, y) subtracting and logarithmically obtaining an image; μ denotes f b (x, y) and f d The value of the connection strength of (x, y) in this embodiment is 0.25; symbol · denotes multiplication; * Represents a convolution;
C. processing the filtered image by X, Y and W channels respectively to obtain X channel contour response, Y channel contour response and W channel contour response;
the formula for obtaining the X channel contour response is as follows:
FR X⌒C (x，y)＝E X⌒C (x，y)-α X⌒C ·T X⌒C (x，y) (4)
wherein E is X⌒C (X, y) is X channelSuppression of the classical receptive field, T X⌒C (X, y) is the inhibitory response of the X-channel nonclassical receptive field, α X⌒C The inhibition intensity is shown.
The inhibition response E for obtaining the X channel classical receptive field X⌒C The process of (x, y) is as follows:
a. the classical receptive field of X cells was simulated using the first derivative of gaussian with directional selectivity:
wherein, the rotation angle theta represents the direction parameter theta epsilon [0, pi ] of the receptive field, gamma represents the constant of the ratio of the long axis and the short axis of the receptive field, and gamma =0.5;
b. let σ = σ XC Obtaining the energy response E of the classical receptive field of the X cells to the external stimulus X⌒C (x，y；θ i )：
E X⌒C (x，y；θ i )＝f C (x，y)*RF(x，y；θ i ，σ XC ) (8)
Wherein C belongs to { g, d }, g represents a gray level image, and d represents a parallax image;
theta as above i Each pixel point is represented as N θ The direction angle in each direction is calculated as follows:
c. comparing the energy information E of each pixel point in different direction angles X⌒C (x，y；θ i ) (ii) a Is obtained whenOptimal response E of the front position X⌒C (x, y) obtaining an inhibitory response of the classical receptive field; simultaneously acquiring corresponding optimal direction angle theta X⌒g (x，y)：
E X⌒C (x，y)＝max{E X⌒C (x，y；θ i )|i＝1，2，…，N θ } (10)
Obtaining the inhibition response E of the X channel non-classical receptive field X⌒C The process of (x, y) is as follows:
a. optimal energy response and suppression weight omega of each pixel point in classical receptive field u (x, y) convolution to obtain the inhibitory response T of the non-classical receptive field X⌒C (x，y)：
T X⌒C (x，y)＝E X⌒C (x，y)*ω n (x，y) (12)
Wherein | | xi | purple 1 Represents L 1 Norm, H (Z) = max (0, Z);
DOG is a two-dimensional gaussian difference function, and its formula is:
wherein σ u Standard deviation of the non-classical receptive field is indicated, K represents the scale ratio of the peripheral inhibitory region to the central excitatory region, K =4 in this example.
In the step C, the formula for obtaining the Y channel contour response is:
FR Y⌒C (x，y)＝S Y⌒C (x，y)-α Y⌒C ·T Y⌒C (x，y) (15)
wherein S is Y⌒C (x, Y) is the Y-channel nonlinear quench response, T Y⌒C (x, Y) is the Y-channel linear suppression response, α Y⌒C The inhibition intensity is shown.
The obtained Y channel nonlinear suppression response T X⌒C The process of (x, y) is as follows:
a. by sigma Y Substitution of σ in equations 1-3, 5, 6 X Obtaining the classical receptive field response E of the Y cell Y⌒C (x，y；θ i ) Then E is added Y⌒C (x，y；θ i ) Substituting into formula 9-11 to obtain the optimal response E of Y cell Y⌒C (x, y) and corresponding optimal direction angle θ y⌒g (x，y)；
b. Setting a region S of finite size xy Each pixel point in the region corresponds to the central position of one subunit, the rectification property of the subunit is simulated by a Gaussian function, and then the nonlinear subunit response S is obtained by summing the contribution of each subunit Y⌒C (x, y) in the formula:
wherein the content of the first and second substances,
representing a window S centered at (x, y) xy Intrinsic response E Y_C Information in (x, y). Sigma s ＝σ YC Calculating the spatial standard deviation of the subunit size; s xy Radius of (r) s ＝12·σ YC (ii) a N is
The number of pixels of (a); v represents the ordinate of the Y-type cell subunit;
obtaining a Y-channel linear inhibitory response T Y⌒C The process of (x, y) is as follows: will S Y⌒C (x, y) instead of E X⌒C (x, y) as inputAnd substituting the formula 12 for calculation.
In the step C, the calculation process for obtaining the W channel contour response is as follows:
let σ = σ n - Will σ n Substituting into formula 5 to obtain first derivative function RF (x, y; theta, sigma) of Gaussian with different scales and shapes n ) (ii) a Standard deviation sigma n The calculation formula of (a) is as follows:
σ WC is the standard deviation of CRF for W cells, since W cells are similar to CRF for X cells, here σ WC ＝σ XC (ii) a S represents the size of the scale change of the classical receptive field of the W cell, and S is 0.2 in the embodiment; n denotes the number of classical receptive field scale changes, where n =25/σ WC ；
b. From a Gaussian first derivative function RF (x, y; theta, sigma) n ) Selecting different micro-motion areas, wherein the selection formula of the micro-motion areas is as follows:
RF (x, y; theta, sigma) n ) Multiplying by equation 19 yields a gaussian first derivative function with fixation and micromotion characteristics:
RM n (x，y；θ i ，σ n )＝RF(x，y；θ i ，σ n )·M n (x，y) (20)；
c. the response of the classical receptive field under the visual micromotion to the external stimulus is input from the characteristic diagram and RM n (x，y；θ i ，σ n ) Convolution yields an energy response E W⌒C (x，y；θ i ，σ n )：
E W⌒C (x，y；θ i ，σ n )＝f C (x，y)*RM n (x，y；θ i ，σ n ) (21)；
d. From E W⌒C (x，y；θ i ，σ n ) In place of E X⌒C (x，y；θ i ) Substituting into the formula 9-11 to calculate the optimal response E of W cells W⌒C (x, y) and corresponding optimal direction angle θ W⌒g (x，y)；
From E W⌒C (x, y) instead of E X⌒C (x, y) as input into equation 12 to calculate the non-classical receptor suppression response T of W cells W⌒C (x，y)；
e. W cell contour response FR W⌒C The formula for (x, y) is:
FR W⌒C (x，y)＝E W⌒C (x，y)-α W⌒C ·T W⌒C (22)
wherein E is W⌒C (x, y) is the optimal response of W cells, T X⌒C (x, y) is the inhibitory response of the W channel non-classical receptive field, α W⌒C The inhibition intensity is shown.
D. Respectively suppressing the X channel contour response, the Y channel contour response and the W channel contour response through sparse coding, removing redundant textures, and obtaining suppressed X channel contour response, Y channel contour response and W channel contour response; the sparse coding formula is as follows:
wherein e is M (x，y)∈{FR M⌒g (X, Y) }, namely M belongs to { X, Y, W } and represents the stimulation response under X, Y and W channels based on the gray level image information;
the histogram of the local gradient amplitude values of each information channel with (x, y) as the center; t is
In this example, a 25 × 25 square window; i 1 Is L 1 Norm, | · | luminance 2 Is L 2 A norm;
represents->
Min represents the minimum value of the two;
applying sparse coding to inhibit, and obtaining formulas of inhibited X channel contour response, Y channel contour response and W channel contour response as follows;
LR M⌒g (x，y)＝sm M (x，y)·FR M⌒g (x，y) (24)。
E. summing the suppressed X-channel contour response, Y-channel contour response and W-channel contour response with the parallax contour response of the corresponding channel of the parallax image respectively to obtain the X-channel contour response, Y-channel contour response and W-channel contour response with strengthened weak contours;
respectively matching the suppressed X-channel contour response, Y-channel contour response and W-channel contour response with the parallax contour response FR of the corresponding channel of the parallax image M⌒d The formula for the sum (x, y) is:
R M ＝FR M⌒d (x，y)+LR M⌒g (x，y) (25)。
F. as shown in fig. 1, fusing the X-channel contour response, the Y-channel contour response, and the W-channel contour response, which strengthen the weak contour, by using different gaze planes to obtain a final contour and an optimal direction;
the process of fusing the X channel contour response, the Y channel contour response and the W channel contour response which strengthen the weak contour by using different watching planes comprises the following steps:
a. the visual information of each watching plane is processed by an X channel, the visual information of a near plane is processed by a Y channel, the visual information of a far plane is processed by a W channel, and the confirmation formula of the watching planes is as follows:
b. normalizing the parallax image to [0, P]Parallax range of (1), using d (x, y) represents the parallax image after normalization; setting the parallax range on the gazing plane as beta, wherein beta is set as 2 in the embodiment; when round (G (x, y), 1) =0.5, it means a near plane; when round (G (x, y), 1) = -0.5 is set, a far plane is represented; setting G (x, y) to other values, indicating on the gaze plane;
c. and fusing visual information on X, Y and W channels through three gazing planes to obtain a final contour R (X, Y) and an optimal direction theta (X, Y):
G. and carrying out non-maximum suppression on the final contour response by using the optimal direction to obtain a suppressed final contour.
2. Comparing the contour identification performance tests based on the method:
1. the method of document 1 was used for comparison:
document 1: lin C, zhang Q, cao Y. Multi-scale control detection model based on quantitative on temporal movement mechanism [ J ]. Signal, image and Video Processing, 2019.
The parameters used in document 1 are the optimal parameters of the model, as in the original text.
2. For quantitative performance evaluation of the final profile, we used the same performance measurement criteria as in document 1, specifically evaluated as follows:
wherein P represents precision and R represents recall. The larger the value of F, the better the performance.
The parameters used in document 1 are the optimal parameters of the model, as in the original text.
The comparative test results are shown in FIG. 2:
FIG. 2 shows four natural images randomly selected by the NYUD data set, a real profile, an optimal profile detected by the method of document 1 and an optimal profile of the invention; among them, F-score is shown in the lower right corner of FIG. 2, and the present invention circled in a red oval in the figure is better in texture suppression than document 1 in this region.
From the experimental effect, the detection method of the invention is superior to the detection method of the document 1.
Claims (10)
1. A contour detection method for simulating the fusion of XYW cell receptive field characteristics is characterized by comprising the following steps:
A. acquiring a gray image and a parallax image of an image to be detected;
B. carrying out rapid bilateral filtering on the gray level image of the image to be detected to obtain a filtered image;
C. processing the filtered image by X, Y and W channels respectively to obtain X channel contour response, Y channel contour response and W channel contour response;
D. respectively suppressing the X channel contour response, the Y channel contour response and the W channel contour response through sparse coding, removing redundant textures, and obtaining suppressed X channel contour response, Y channel contour response and W channel contour response;
E. respectively summing the suppressed X-channel contour response, Y-channel contour response and W-channel contour response with the parallax contour response of the corresponding channel of the parallax image to obtain the X-channel contour response, Y-channel contour response and W-channel contour response with strengthened weak contours;
F. fusing the X channel contour response, the Y channel contour response and the W channel contour response of the strengthened weak contour by using different watching planes to obtain a final contour and an optimal direction;
G. and carrying out non-maximum suppression on the final contour response by using the optimal direction to obtain a suppressed final contour.
2. The method of claim 1 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
in step B, the formula of the fast bilateral filtering is:
f g (x，y)＝exp(μ· - f b (x，y)+f d (x，y) (1)
f d (x，y)＝log(I(x，y)-f b (x，y)) (3)
where I (x, y) represents a gray image, σ represents a standard deviation of a Gaussian function, and f b (x, y) represents an image of the gray image after the Gaussian function smoothing, f d (x, y) is represented by a gray image and f b (x, y) is a difference and logarithmic, and μ represents f b (x, y) and f d The (x, y) link strength, sign · represents multiplication, and one represents convolution.
3. The method of claim 1 for contour detection that mimics the fusion of receptor field properties of a XYW cell, comprising:
in the step C, the formula for obtaining the X channel contour response is:
wherein the content of the first and second substances,
is a suppressive response of the X-channel classical receptive field, is->
Is a suppressive response of the X-channel non-classical receptive field, is based on the presence of a receptor in the receptor>
The inhibition intensity is shown.
4. The method of claim 3 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
obtaining the inhibition response of the classical receptive field of the X channel
The calculation process comprises the following steps:
a. the classical receptive field of X cells was simulated using the first derivative of gaussian with directional selectivity:
wherein, the rotation angle theta represents the direction parameter theta epsilon [0, pi ] of the receptive field, gamma represents the constant of the ratio of the long axis and the short axis of the receptive field, and gamma =0.5;
b. let σ = σ XC Obtaining the energy response of the classical receptive field of the X cells to the external stimulus
C belongs to { g, d }, g represents a gray level image, and d represents a parallax image;
theta as above i Each pixel point is represented as N θ The direction angle in each direction is calculated as follows:
c. comparing the energy information of each pixel point in different direction angles
Obtaining an optimal response to the current location->
Obtaining the inhibition response of the classical receptive field; simultaneously acquiring corresponding optimal direction angles
Obtaining inhibition response of X-channel non-classical receptive field
The calculation process comprises the following steps:
a. optimal energy response and suppression weight omega of each pixel point in classical receptive field u (x, y) convolution to obtain the inhibitory response of the non-classical receptive field
Wherein | | xi | purple 1 Represents L 1 Norm, H (Z) = max (0, Z);
DOG is a two-dimensional gaussian difference function, and its formula is:
wherein σ u Standard deviation of the non-classical receptive field is indicated and k represents the scale ratio of the peripheral inhibitory region to the central excitatory region.
5. The method of claim 4 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
in the step C, the formula for obtaining the Y channel contour response is:
wherein the content of the first and second substances,
for a Y-channel non-linear suppression response, a->
For a Y-channel linear suppression response, a->
The inhibition intensity is indicated.
6. The method of claim 5 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
obtaining Y-channel nonlinear rejection response
The calculation process comprises the following steps:
a. by sigma Y Substitution of σ in equations 1-3, 5, 6 X Obtaining the classical receptor field response of Y cells
Then will
Substitution into equations 9-11 results in an optimal response for Y cells>
And corresponding optimal direction angle
b. Setting a region S of finite size xy Each pixel point in the region corresponds to the central position of one subunit, the rectification property of the subunit is simulated by a Gaussian function, and then the nonlinear subunit response is obtained by summing the contribution of each subunit
The formula is as follows:
wherein the content of the first and second substances,
representing a window S centered on (x, y) xy Intrinsic response E Y_C Information in (x, y), σ s ＝σ YC Calculating the spatial standard deviation of the size of the subunit; s xy Radius of (d) size r s ＝12·σ YC (ii) a N is->
The number of pixels of (a); v represents the ordinate of the Y-cell subunit;
the linear suppressive response of the acquisition Y channel
The calculation process comprises the following steps: will->
Instead of the former
And is calculated by substituting the formula 12 as an input.
7. The method of claim 4 for contour detection that mimics fusion of receptor field properties of a XYW cell, characterized in that:
in the step C, the calculation process of obtaining the W channel contour response is:
let σ n = σ n - Will σ n Substituting into formula 5 to obtain first derivative function RF (x, y; theta, sigma) of Gaussian with different scales and shapes n ) (ii) a Standard deviation sigma n The calculation formula of (a) is as follows:
σ WC is the standard deviation of CRF for W cells, since W cells are similar to CRF for X cells, here σ WC ＝ XC (ii) a S represents the size of the scale change of the classical receptive field of the W cell; n denotes the number of classical receptive field scale changes, where n =25/σ WC ；
b. From a Gaussian first derivative function RF (x, y; θ i, σ) n ) Selecting different micro-motion areas, wherein the selection formula of the micro-motion areas is as follows:
RF (x, y; theta, sigma) n ) Multiplying by equation 19 yields a gaussian first derivative function with fixation and fixation micromotion characteristics:
RM n (x，y；θ i ，σ n )＝RF(x，y；θ i ，σ n )·M n (x，y) (20)；
c. the response of the classical receptive field under the visual micromotion to the external stimulus is input from the characteristic diagram and RM n (x，y；θ i ，σ n ) Convolution yields an energy response
d. By
In place of>
Substituting into equation 9-11 to calculate the optimal response of W cells
And the corresponding optimum direction angle->
/>
By
In place of>
Substituting as input into equation 12 calculates the non-canonical receptive field suppression response +for W cells>
e. W cell contour response
The calculation formula is as follows:
wherein the content of the first and second substances,
for optimal response of W cells>
In response to suppression of the W channel non-classical receptive field>
The inhibition intensity is shown.
8. The method of claim 1 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
the sparse coding formula in the step D is as follows:
wherein the content of the first and second substances,
that is, M belongs to { X, Y, W } represents the stimulation response under X, Y and W channels based on the gray image information; />
Is a histogram of local gradient amplitude values of each information channel centered at (x, y), t is
Dimension of (d); i | · | live through 1 Is L 1 Norm, | · | luminance 2 Is L 2 A norm; />
To represent
Min represents the minimum value of the two;
applying sparse coding to inhibit, and obtaining formulas of inhibited X channel contour response, Y channel contour response and W channel contour response as follows;
9. the method of claim 1 for contour detection that mimics the fusion of receptor field properties of a XYW cell, comprising:
in the step E, the suppressed X-channel contour response, Y-channel contour response and W-channel contour response are respectively compared with the parallax contour response of the corresponding channel of the parallax image
The formula for the summation is:
10. the method of claim 1 for contour detection that mimics the fusion of receptor field properties of XYW cells, comprising:
in the step F, the process of fusing the X channel contour response, the Y channel contour response and the W channel contour response which strengthen the weak contour by using different gazing planes is as follows:
a. the visual information of each watching plane is processed by an X channel, the visual information of a near plane is processed by a Y channel, the visual information of a far plane is processed by a W channel, and the confirmation formula of the watching planes is as follows:
b. normalizing the parallax image to [0, P]Parallax range of (1), using d (x, y) represents the parallax image after normalization; setting the parallax range on the gazing plane as beta; when round (G (x, y), 1) =0.5, it represents a near plane; when round (G (x, y), 1) = -0.5 is set, a far plane is represented; setting G (x, y) to other values, indicating on the gaze plane;
c. and fusing visual information on X, Y and W channels through three gazing planes to obtain a final contour R (X, Y) and an optimal direction theta (X, Y):
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110493954.0A CN113192092B (en) | 2021-05-07 | 2021-05-07 | Contour detection method for simulating fusion of properties of receptor field of XYW cell |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110493954.0A CN113192092B (en) | 2021-05-07 | 2021-05-07 | Contour detection method for simulating fusion of properties of receptor field of XYW cell |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113192092A CN113192092A (en) | 2021-07-30 |
| CN113192092B true CN113192092B (en) | 2023-04-07 |
Family
ID=76983981
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110493954.0A Active CN113192092B (en) | 2021-05-07 | 2021-05-07 | Contour detection method for simulating fusion of properties of receptor field of XYW cell |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113192092B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117079397B (en) * | 2023-09-27 | 2024-03-26 | 青海民族大学 | Wild human and animal safety early warning method based on video monitoring |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107067407A (en) * | 2017-04-11 | 2017-08-18 | 广西科技大学 | Profile testing method based on non-classical receptive field and linear non-linear modulation |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4803736A (en) * | 1985-11-27 | 1989-02-07 | The Trustees Of Boston University | Neural networks for machine vision |
| US6941016B1 (en) * | 2001-12-31 | 2005-09-06 | Cognex Technology And Investment | Method for finding contours in an image of an object |
| US9092692B2 (en) * | 2012-09-13 | 2015-07-28 | Los Alamos National Security, Llc | Object detection approach using generative sparse, hierarchical networks with top-down and lateral connections for combining texture/color detection and shape/contour detection |
| CN109919945B (en) * | 2019-02-01 | 2022-03-25 | 广西科技大学 | Contour detection method based on non-classical receptive field non-linear two-side subunit response |
| CN111179294B (en) * | 2019-12-30 | 2020-09-15 | 广西科技大学 | Bionic type contour detection method based on X, Y parallel visual channel response |
| CN111968140B (en) * | 2020-06-24 | 2023-07-14 | 广西科技大学 | Contour detection method based on classical receptive field vision-strengthening micro-motion mechanism |
| CN111968141B (en) * | 2020-06-30 | 2023-06-16 | 广西科技大学 | Non-classical receptive field contour detection method based on multi-scale superposition |
-
2021
- 2021-05-07 CN CN202110493954.0A patent/CN113192092B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107067407A (en) * | 2017-04-11 | 2017-08-18 | 广西科技大学 | Profile testing method based on non-classical receptive field and linear non-linear modulation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113192092A (en) | 2021-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Low-illumination image enhancement algorithm based on improved multi-scale Retinex and ABC algorithm optimization | |
| CN109117826B (en) | Multi-feature fusion vehicle identification method | |
| Zhou et al. | Utilizing dictionary learning and machine learning for blind quality assessment of 3-D images | |
| WO2015180527A1 (en) | Image saliency detection method | |
| CN111402285B (en) | Contour detection method based on visual mechanism dark edge enhancement | |
| CN114926459B (en) | Image quality evaluation method, system and computer readable medium | |
| CN110210493B (en) | Contour detection method and system based on non-classical receptive field modulation neural network | |
| Saif et al. | Gradient based image edge detection | |
| CN109788275A (en) | Naturality, structure and binocular asymmetry are without reference stereo image quality evaluation method | |
| CN105046202B (en) | Adaptive recognition of face lighting process method | |
| CN116385315B (en) | Image enhancement method and system for simulated ablation of tumor therapeutic instrument | |
| Dash et al. | Illumination normalized based technique for retinal blood vessel segmentation | |
| CN113192092B (en) | Contour detection method for simulating fusion of properties of receptor field of XYW cell | |
| Gu et al. | Research on the improvement of image edge detection algorithm based on artificial neural network | |
| CN110555820A (en) | Image fusion method based on convolutional neural network and dynamic guide filtering | |
| Madrid-Herrera et al. | Human image complexity analysis using a fuzzy inference system | |
| CN113129390B (en) | Color blindness image re-coloring method and system based on joint significance | |
| CN115187569A (en) | Image quality evaluation model training method, device, equipment and medium | |
| CN114066812B (en) | No-reference image quality evaluation method based on spatial attention mechanism | |
| CN109376782B (en) | Support vector machine cataract classification method and device based on eye image features | |
| CN114982846B (en) | Control method and device suitable for coffee roasting and electronic equipment | |
| CN115187519B (en) | Image quality evaluation method, system and computer readable medium | |
| CN107024480A (en) | A kind of stereoscopic image acquisition device | |
| CN111968140B (en) | Contour detection method based on classical receptive field vision-strengthening micro-motion mechanism | |
| Kumar et al. | Signal and image processing in medical applications |
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 | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20210730 Assignee: HUALI FAMILY PRODUCTS CO.,LTD. Assignor: GUANGXI University OF SCIENCE AND TECHNOLOGY Contract record no.: X2023980054119 Denomination of invention: A contour detection method for simulating the fusion of receptive field characteristics of XYW cells Granted publication date: 20230407 License type: Common License Record date: 20231226 |
|
| EE01 | Entry into force of recordation of patent licensing contract |