CN111539969A - Image edge detection method and device, computer equipment and storage medium - Google Patents

Image edge detection method and device, computer equipment and storage medium Download PDF

Info

Publication number
CN111539969A
CN111539969A CN202010327742.0A CN202010327742A CN111539969A CN 111539969 A CN111539969 A CN 111539969A CN 202010327742 A CN202010327742 A CN 202010327742A CN 111539969 A CN111539969 A CN 111539969A
Authority
CN
China
Prior art keywords
energy response
preset
edge detection
pixel point
image
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
Application number
CN202010327742.0A
Other languages
Chinese (zh)
Other versions
CN111539969B (en
Inventor
肖洁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Railway Vocational College of Technology
Original Assignee
Wuhan Railway Vocational College of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Wuhan Railway Vocational College of Technology filed Critical Wuhan Railway Vocational College of Technology
Priority to CN202010327742.0A priority Critical patent/CN111539969B/en
Publication of CN111539969A publication Critical patent/CN111539969A/en
Application granted granted Critical
Publication of CN111539969B publication Critical patent/CN111539969B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention is suitable for the technical field of computers, and provides an image edge detection method, an image edge detection device, computer equipment and a storage medium, wherein the method comprises the following steps: determining the response results of each pixel point in a plurality of response orientations according to the visual neuron energy response model, calculating and determining the direction selection significance of each pixel point, correcting the scale parameters in the visual neuron energy response model according to the size relation between the direction selection significance and the direction selection significance threshold, re-determining the optimal energy response result and the optimal response orientation of each pixel point, and then outputting the edge detection result by combining with a non-classical receptive field edge detection algorithm. According to the image edge detection method provided by the invention, after the direction selection significance of each pixel point is determined, the scale parameter is corrected according to the direction selection significance, and the response models with different scales are adopted to pertinently extract the outline or inhibit the texture of each pixel point, so that the detection effect in the edge detection result is effectively improved.

Description

Image edge detection method and device, computer equipment and storage medium
Technical Field
The invention belongs to the technical field of computers, and particularly relates to an image edge detection method and device, computer equipment and a storage medium.
Background
In the process of processing an image by using a computer vision technology, an edge detection algorithm is generally needed to distinguish a meaningful target edge from a meaningless background texture in the image, wherein how to better extract the meaningful target edge from the image and suppress the meaningless background texture in the image is an important research direction of many scholars. Biological research finds that in the classical receptive field of human brain visual cortex neurons, a larger area of non-classical receptive field exists, so that the human brain visual cortex can easily distinguish target edges from background textures in an image.
At present, the edge detection algorithm with the best edge extraction effect in the prior art is based on a non-classical receptive field edge detection algorithm, and a novel spatial unified modulation operator is constructed to obtain the enhancement and inhibition influence of peripheral stimulation on central neurons, so that background textures irrelevant to a target are effectively inhibited, and an image edge contour is kept as far as possible. However, the edge detection algorithm still has the problems of insignificant extraction of the image edge contour and missing of the image edge contour.
Disclosure of Invention
The embodiment of the invention aims to provide an image edge detection algorithm, and aims to solve the technical problems of the existing edge detection algorithms that the image edge contour extraction is not obvious, the image edge contour is missing and the like, and the detection effect is not good.
The embodiment of the invention is realized in such a way that an image edge detection method comprises the following steps:
determining energy response results of pixel points in the image to be processed in a plurality of preset response orientations according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
according to the magnitude relation between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, correcting the scale parameter in the visual neuron energy response model to generate a corrected visual neuron energy response model;
determining the optimal energy response result and the optimal response direction of the pixel points according to the corrected visual neuron energy response model;
and determining an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
Another object of an embodiment of the present invention is to provide an image edge detecting apparatus, including:
the energy response result calculation module is used for determining energy response results of pixel points in the image to be processed in a plurality of preset response directions according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
the direction selection significance calculation module is used for determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
the scale parameter correction module is used for correcting the scale parameter in the visual neuron energy response model according to the size relationship between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, and generating a corrected visual neuron energy response model;
the optimal energy response result and orientation calculation module is used for determining the optimal energy response result and the optimal response orientation of the pixel points according to the corrected visual neuron energy response model;
and the edge detection result determining module is used for determining the edge detection result of the image to be processed according to the non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
It is a further object of an embodiment of the present invention to provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, causes the processor to execute the steps of the image edge detection method as described above.
It is another object of an embodiment of the present invention to provide a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, causes the processor to execute the steps of the image edge detection method as described above.
The image edge detection method provided by the embodiment of the invention determines the energy response results of pixel points in an image to be processed on a plurality of preset response orientations according to a visual neuron energy response model, further calculates the direction selection significance of the pixel points, judges whether the pixel points are direction significant pixel points which have significance in a certain orientation and are more likely to represent important structures such as a target contour or non-direction significant pixel points which are more likely to represent non-important structures such as texture features and the like by using the magnitude relation between the direction selection significance and a preset direction selection significance threshold value, corrects the visual neuron energy response model according to different scale parameter correction rules aiming at the direction significant pixel points and the non-direction significant pixel points, and respectively calculates the optimal energy response result and the optimal response result of the corresponding pixel points by using the corrected visual neuron energy response model And finally determining an edge detection result by combining a non-classical receptive field edge detection algorithm. Compared with the prior art, the image edge detection method provided by the invention directly utilizes the same visual neuron energy response model to determine the optimal energy response result and the optimal response direction of each pixel point, the visual neuron energy response model is adjusted according to the direction selection significance of the pixel points, different visual neuron energy response models with scale parameters are used for the pixel points with different properties, the determined optimal energy response result and optimal response direction have better effect when being used for a non-classical receptive field edge detection algorithm, and the inhibition effect of textures in the edge detection result and the extraction effect of outlines are better.
Drawings
FIG. 1 is a flowchart illustrating steps of a method for detecting an edge of an image according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a step of determining an edge detection result according to a non-classical receptive field edge detection algorithm according to an embodiment of the present invention;
fig. 3 is a flowchart illustrating a step of determining a boundary pixel according to an embodiment of the present invention;
FIG. 4 is a flowchart of a step of determining local contrast according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a step of modifying a radial parameter of a receptive field according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating a step of modifying a scale parameter according to an embodiment of the present invention;
FIG. 7 is a schematic structural diagram of an apparatus for detecting an edge of an image according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a computer device capable of executing an image edge detection method according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In order to further improve the detection effect of the existing image edge detection method on the image edge, after the energy response results of the pixel points in the image to be processed in a plurality of preset response orientations are determined according to the visual neuron energy response model, the direction selection significance of each pixel point in the image is determined by further utilizing the preset direction selection significance calculation rule, so that the fact that the energy response results of the pixel points in the image to be processed in the plurality of preset response orientations are determined by directly determining the preset visual neuron energy response model of the pixel points in the image to be processed according to the preset visual neuron energy response model compared with the existing image edge detection method, whether the direction significance pixel points which have significance in a certain orientation and are more likely to represent important structures such as a target contour or the like or the direction significance pixel points which are more likely to represent important structures such as the target contour are relatively balanced in each orientation response is judged, The method is characterized in that the method further considers that a good effect can be obtained by inhibiting the texture on a fine scale, a good effect can be obtained by determining the orientation of a target contour on a coarse scale, the scale parameters of the visual neuron energy response model are respectively adjusted in an adaptive mode aiming at pixels with different properties to determine the optimal energy response result and the optimal response orientation of the pixel, and compared with the conventional image edge detection method which directly utilizes the same visual neuron energy response model to determine the optimal energy response result and the optimal response orientation of each pixel, the method can achieve a better inhibiting effect or an extracting effect on the pixel by combining a non-classical receptive field edge detection algorithm, so that a better edge detection result is obtained.
As shown in fig. 1, a flowchart of steps of an image edge detection method provided in an embodiment of the present invention specifically includes the following steps:
and S102, determining energy response results of pixel points in the image to be processed on a plurality of preset response orientations according to a preset visual neuron energy response model.
In the embodiment of the present invention, the preset visual neuron energy response model may be a visual neuron energy response model determined in advance through machine learning, wherein the visual neuron energy response model should include a scale parameter and also a response orientation parameter. Specifically, the energy response result of each pixel point in the image to be processed can be determined by performing convolution processing on the image to be processed by using the visual neuron energy response model, and when the response orientation parameter in the visual neuron energy response model is changed, the energy response result of each pixel point in the image to be processed in a plurality of preset response orientations can be obtained.
In a preferred embodiment of the present invention, considering that the Gabor filter and the LogGabor filter exhibit significant directional selectivity, preferably, the visual neuron energy response model is established in advance based on the Gabor filter or the LogGabor filter, and may also exhibit significant directional selectivity in a process of processing an image to be processed, so as to ensure that energy response results of each pixel point in different corresponding orientations have significant characteristics.
In the embodiment of the present invention, the visual neuron energy response model established in advance based on the Gabor filter is specifically as follows:
Figure BDA0002463819120000061
wherein, the meaning of a plurality of parameters contained in the visual neuron energy response model is specifically as follows: λ is the wavelength, 1/λ is the spatial frequency of the cosine function; sigma is the standard deviation of a Gaussian function of the size of the elliptical receptive field, and sigma/lambda represents the bandwidth of the spatial frequency, namely a scale parameter, wherein the larger the bandwidth of the spatial frequency is, the larger the scale of the visual neuron energy response model is, and the smaller the bandwidth of the spatial frequency is, the smaller the scale of the visual neuron energy response model is; theta is a neuron preference direction, namely the response orientation parameter, and the energy response results of pixel points in the image to be processed on a plurality of preset response orientations (theta) can be determined by adjusting theta; gamma is the ratio of the length to the length of the oval receptive field;
Figure BDA0002463819120000062
is a phase parameter;
Figure BDA0002463819120000063
Figure BDA0002463819120000064
at this time, the formula for performing convolution processing on the gray-scale image I (x, y) of the image to be processed according to the visual neuron energy response model is specifically as follows:
Figure BDA0002463819120000065
wherein, the mode of Gabor filter with phase parameter difference of pi/2 to image response is energy response result E for describing basic characteristics of each pixel pointλ,σ,θ,γ(x, y), specifically:
Figure BDA0002463819120000066
it can be seen that Eλ,σ,θ,γ(x, y) is
Figure BDA0002463819120000067
And
Figure BDA0002463819120000068
the two Gabor filters differ in phase parameter by pi/2 modulo the response of the image.
Further, let NθA preset response direction thetai(i.e., the neuron preference direction parameter θ of the visual neuron energy response model) is:
Figure BDA0002463819120000071
at the moment, energy response results of pixel points in the image to be processed on a plurality of preset response directions can be determined
Figure BDA0002463819120000072
In the embodiment of the present invention, the energy response model of the visual neuron established in advance based on the LogGabor filter is specifically as follows:
Figure BDA0002463819120000073
similarly, the meaning of the plurality of parameters included in the visual neuron energy response model is specifically as follows: f. of0Is the filter center frequency; sigmafFor radial bandwidth, σθThe directional bandwidth is a scale parameter, the larger the bandwidth of the radial bandwidth and the directional bandwidth is, the larger the scale of the visual neuron energy response model is, and the smaller the bandwidth of the radial bandwidth and the directional bandwidth is, the smaller the scale of the visual neuron energy response model is; theta0For the filter direction, i.e. the response orientation parameter mentioned above, by adjusting θ0And determining energy response results of pixel points in the image to be processed in a plurality of preset response directions. At the moment, convolution processing is carried out on the gray level image I (x, y) of the image to be processed according to the visual neuron energy response model, and an energy response result E describing the basic characteristics of each pixel point is obtainedf,θThe specific formula of (x, y) is:
Figure BDA0002463819120000074
hoddcorresponding to the real part in the H spatial domain of the filter, even filter, HevenThe imaginary part in the H spatial domain of the corresponding filter is an odd filter.
Further, N may be the sameθA preset response direction thetai(i.e., neuron preference direction parameter θ of visual neuron energy response model0) Comprises the following steps:
Figure BDA0002463819120000075
at the moment, energy response results of pixel points in the image to be processed on a plurality of preset response directions can be determined
Figure BDA0002463819120000076
In the embodiment of the present invention, it should be noted that, in the prior art, after determining the energy response results in the plurality of preset response orientations, the maximum energy response result is directly obtained
Figure BDA0002463819120000081
Or
Figure BDA0002463819120000082
As a result of the optimal energy response, corresponding response orientation
Figure BDA0002463819120000083
Or
Figure BDA0002463819120000084
And as an optimal corresponding direction, bringing the optimal energy response result and the optimal corresponding direction determined by the prior art into a non-classical receptive field edge detection algorithm to perform edge detection calculation.
Figure BDA0002463819120000085
Figure BDA0002463819120000086
Figure BDA0002463819120000087
Figure BDA0002463819120000088
And step S104, determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule.
In the embodiment of the present invention, compared with the prior art in which the optimal energy response result and the optimal response orientation of each pixel point in the image are directly determined, the present invention further determines the direction selection saliency of the pixel point by using the energy response result and a preset direction selection saliency calculation rule, and for convenience of understanding, the foregoing visual neuron energy response model established based on the Gabor filter is used as an example to describe the calculation rule of the direction selection saliency of the pixel point. It should be noted that the calculation rule of the direction selection significance may also be extended to other optic neuron energy response models, such as an optic neuron energy response model established based on a LogGabor filter, which may be made by those skilled in the art according to common knowledge.
In the embodiment of the present invention, the direction selection saliency of a pixel point is determined by calculating the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions, and a specific calculation formula of the direction selection saliency of the pixel point is as follows:
Figure BDA0002463819120000089
the numerator on the right of the equation is the maximum energy response result of the pixel point in a plurality of preset directions, and the denominator is the average energy response result of the pixel point in the plurality of preset directions. It can be understood that the greater the direction selection significance, i.e., the greater the quotient of the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions, the more obvious the response of the pixel point in a certain direction is, and the more likely the pixel point is to be the contour feature of the image, and the smaller the direction selection significance, i.e., the greater the quotient of the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions is, the more likely the pixel point is to be the texture feature of the image.
And step S106, according to the size relationship between the direction selection significance and a preset direction selection significance threshold value, a preset scale parameter correction rule and a scale parameter in the visual neuron energy response model, correcting to generate a corrected visual neuron energy response model.
In the embodiment of the invention, the direction of each pixel point is selected to be the significance Sλ,σ(x, y) and a preset direction selection saliency threshold tsAnd comparing to determine whether the pixel point is a direction significant pixel point which has a particularly obvious response in a certain direction and is more likely to represent important structures such as a target contour and the like, or a non-direction significant pixel point which has a relatively balanced response in each direction and is more likely to represent non-important structures such as texture features and the like.
In the embodiment of the invention, considering that the image is processed by using the fine-scale (i.e. the scale parameter is relatively small) visual neuron energy response model to have a good inhibition effect on the texture, and the image is processed by using the coarse-scale (i.e. the scale parameter is relatively large) visual neuron energy response model to have a good extraction effect on the outline, therefore, after each pixel point is determined to be a direction significant pixel point or a non-direction significant pixel point, the scale parameter contained in the visual neuron energy response model is adaptively corrected, the non-direction significant pixel point is processed by using the fine-scale visual neuron energy response model, and the direction significant pixel point is processed by using the coarse-scale visual neuron energy response model. Please refer to the following fig. 6 and the explanation thereof for a specific step of modifying the scale parameter in the energy response model of the visual neuron.
In the embodiment of the present invention, it is emphasized that the coarse scale and the fine scale referred in the present invention only represent relative concepts, and the magnitude of the scale parameter is not particularly limited, the coarse scale visual neuron energy response model and the fine scale coarse scale visual neuron energy response model can be represented, but the scale parameter of the visual neuron energy response model is modified according to the directional significance of the pixel point, so that the technical scheme of respectively processing the pixel point with (without) directional significance by using the visual neuron energy response models with two or more different sizes of the scale parameter is within the scope claimed in the present invention.
And S108, determining the optimal energy response result and the optimal response direction of the pixel point according to the corrected visual neuron energy response model.
In the embodiment of the present invention, each pixel point is processed based on the visual neuron energy response model after the scale parameter correction, and an optimal energy response result and an optimal response direction of the pixel point are further determined, where the calculation rules of the optimal energy response result and the optimal response direction are described in the foregoing step S102. That is to say, the scale parameters in the visual neuron energy response model processed for each pixel point are corrected based on the direction selection significance of the pixel point, so that a good suppression or extraction effect on each pixel point is ensured.
And step S110, determining an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
In the embodiment of the invention, a non-classical receptive field edge detection algorithm considering the influence of context modulation is utilized, and the optimal energy response result and the optimal response direction of each pixel point in the image are combined, so that the energy response result of each pixel point can be further corrected by utilizing the context, the corrected energy response result of each pixel point is displayed, and the edge detection result of the image to be processed can be obtained.
In the embodiment of the present invention, please refer to fig. 2 and the explanation thereof for the step of determining the edge detection result according to the non-classical receptive field edge detection algorithm.
The image edge detection method provided by the embodiment of the invention determines the energy response results of pixel points in an image to be processed on a plurality of preset response orientations according to a visual neuron energy response model, further calculates the direction selection significance of the pixel points, judges whether the pixel points are direction significant pixel points which have significance in a certain orientation and are more likely to represent important structures such as a target contour or non-direction significant pixel points which are more likely to represent non-important structures such as texture features and the like by using the magnitude relation between the direction selection significance and a preset direction selection significance threshold value, corrects the visual neuron energy response model according to different scale parameter correction rules aiming at the direction significant pixel points and the non-direction significant pixel points, and respectively calculates the optimal energy response result and the optimal response result of the corresponding pixel points by using the corrected visual neuron energy response model And finally determining an edge detection result by combining a non-classical receptive field edge detection algorithm. Compared with the prior art, the image edge detection method provided by the invention directly utilizes the same visual neuron energy response model to determine the optimal energy response result and the optimal response direction of each pixel point, the visual neuron energy response model is adjusted according to the direction selection significance of the pixel points, different visual neuron energy response models with scale parameters are used for the pixel points with different properties, the determined optimal energy response result and optimal response direction have better effect when being used for a non-classical receptive field edge detection algorithm, and the inhibition effect of textures in the edge detection result and the extraction effect of outlines are better.
As shown in fig. 2, a flowchart of the steps for determining the edge detection result according to the non-classical receptive field edge detection algorithm provided in the embodiment of the present invention specifically includes the following steps:
step S202, determining boundary pixel points according to the determined central pixel points and preset receptive field radius parameters.
In the embodiment of the invention, the non-classical receptive field edge detection algorithm is an algorithm discovered based on biological research, and the main principle is that the edge detection of an image is realized by taking the non-classical receptive field with a larger area as a reference in the classical receptive field of human brain cortical neuron. Specifically, in the edge detection process, for each pixel point, not only the optimal energy response result of the pixel point is considered, but also the modulation influence of other pixel points in the receptive field radius on the pixel point is additionally considered, at this time, the pixel point is considered as a central pixel point, and a boundary pixel point can be determined by combining the receptive field radius parameter, wherein the distance between the boundary pixel point and the central pixel point is smaller than the receptive field radius parameter.
As a preferred embodiment of the present invention, the receptive field radius parameter may also be corrected for different central pixel points, so as to achieve a better contour extraction effect and a better texture suppression effect. Please refer to fig. 3 and the explanation thereof for a specific implementation manner.
Step S204, determining the offset angle and the distance between the central pixel point and the boundary pixel point.
In the embodiment of the present invention, the optimal energy response result and the optimal response position of each pixel point are provided in the foregoing steps, and at this time, an angular difference (β - ω) between the optimal response position β of the central pixel point and the optimal response position of the boundary pixel point is calculated, and the angular difference is the angular differenceIs the offset angle phi between the central pixel point and the boundary pixel pointABIt should be noted that by taking the smaller of the angle and the complement of the angle, i.e., +ABThe offset angle phi can be ensured when the offset angle phi is min (| β -omega, pi- | β -omega |)ABBetween 0 and pi/2. The distance can be determined directly by a distance formula, which is not described herein.
Step S206, determining the energy response influence value of the boundary pixel point on the central pixel point according to the offset angle and the distance and the optimal energy response result of the boundary pixel point.
In the embodiment of the present invention, when phiABWhen the value is 0, the enhancement effect of the boundary pixel point on the central pixel point is maximum (meaning that the optimal response direction of the boundary pixel point points to the position of the central pixel point and shows collinear enhancement or angular point enhancement); and phiABWhen the ratio is pi/2, the inhibition effect of the boundary pixel point on the central pixel point is maximum (meaning that the path between the central pixel point and the boundary pixel point is blocked by peripheral neurons and shows texture inhibition or blocking inhibition), so that the influence degree of the offset angle on the boundary pixel point on the central pixel point can be effectively simulated by utilizing a cosine function.
In the embodiment of the invention, similarly, the distance also affects the degree of influence of the boundary pixel points on the central pixel points, and the farther the boundary pixel points are from the central pixel points, the smaller the degree of influence is, and the closer the boundary pixel points are from the central pixel points, the larger the degree of influence is.
In the embodiment of the present invention, in combination with the above, a spatial uniform modulation operator can be defined:
Figure BDA0002463819120000131
at this time, the reference ΩNCRFDenotes the scope of receptive field as rNCRFThe parameter representing the radius of the receptive field is that the calculation formula of the energy response influence value of the boundary pixel point to the central pixel point is specifically as follows:
Figure BDA0002463819120000132
wherein,
Figure BDA0002463819120000133
namely, the distance between the boundary pixel point and the central pixel point is smaller than the receptive field radius parameter.
And S208, determining the complete energy response of the central pixel point according to the energy response influence value and the optimal energy response of the central pixel point.
In the embodiment of the invention, the influence of context modulation is controlled by taking a weighted value for the energy response influence value, and the optimal energy response of the central pixel point and the energy response influence value of the boundary pixel point to the central pixel point are synthesized, so that the complete energy response of the central pixel point after considering the context modulation mechanism can be determined by summing.
And step S210, determining an edge detection result of the image to be processed according to the complete energy response.
In the embodiment of the invention, each pixel point is taken as a central pixel point, the complete energy response of each pixel point is calculated and displayed, and the edge detection result of the image to be processed can be obtained.
According to the image edge detection method provided by the embodiment of the invention, the non-classical receptive field edge detection algorithm is further utilized, and the context modulation influence of the surrounding environment is comprehensively considered, so that the processing effect of image edge detection is further improved.
As shown in fig. 3, a flowchart of a step of determining a boundary pixel point provided in an embodiment of the present invention specifically includes the following steps:
step S302, determining the local contrast of the central pixel point according to the gray image of the image to be processed.
In the embodiment of the present invention, the local contrast of the central pixel point may be determined by the maximum brightness value and the minimum brightness value of the image in the receptive field range, wherein the specific calculation process refers to fig. 4 and the explanation thereof.
Step S304, the receptive field radius parameter is corrected according to the magnitude relation between the local contrast and a preset low-contrast response threshold and a preset receptive field radius parameter correction rule, and a corrected receptive field radius parameter is generated.
In the embodiment of the present invention, when the local contrast of the central pixel is too low, the spatial action range of the non-classical receptive field is changed by modifying the receptive field radius parameter to improve the enhancing and suppressing effects of the model, wherein the specific modification rule please refer to fig. 5 and the content explained in the description thereof.
As a preferred embodiment of the present invention, an average value of local contrast of each pixel point in an image is determined as the preset low-contrast response threshold, that is:
Figure BDA0002463819120000141
wherein, CtFor the low contrast response threshold, M, N is the number of pixels in the X, Y direction of the input image, and C (x, y) represents the local contrast when (x, y) is the central pixel.
And S306, determining boundary pixel points according to the central pixel points and the corrected receptive field radius parameters.
In the embodiment of the invention, when different pixel points are used as central pixel points, the receptive field radius parameter is corrected according to the local contrast of the receptive field range so as to enlarge (reduce) the spatial action range of the receptive field.
As shown in fig. 4, a flowchart of a step of determining a local contrast provided in an embodiment of the present invention specifically includes the following steps:
and S402, determining a gray image in the receptive field range of the central pixel point according to a preset receptive field radius parameter.
In the embodiment of the present invention, obviously, with the central pixel point as a center, the pixel points whose distance from the central pixel point is smaller than the preset receptive field radius parameter all belong to the receptive field range of the central pixel point, and at this time, the gray level image in the receptive field range of the corresponding central pixel point can be determined.
Step S404, determining the maximum brightness value and the minimum brightness value of the gray level image in the receptive field range.
In the present embodiment, Lmax、LminRespectively representing the maximum brightness value and the minimum brightness value of the gray-scale image in the receptive field range.
Step S406, determining the local contrast of the central pixel point according to the maximum brightness value and the minimum brightness value.
In the embodiment of the present invention, the calculation formula of the local contrast is as follows:
Figure BDA0002463819120000151
as shown in fig. 5, a flowchart of a step of correcting a receptive field radius parameter provided in an embodiment of the present invention specifically includes the following steps:
step S502, judging whether the local contrast is smaller than a preset low contrast response threshold value. When the local contrast is judged to be smaller than a preset low-contrast response threshold, executing step S504; and executing other steps when the local contrast is judged to be smaller than a preset low-contrast response threshold value.
In the embodiment of the invention, when the local contrast is smaller than the preset low-contrast response threshold, the brightness of the pixel point is relatively balanced, and the spatial action range of the receptive field is expanded by expanding the radius parameter of the receptive field, so that the enhancement and inhibition effects of the model are improved. When the local contrast is not less than the preset low-contrast response threshold, it indicates that the brightness of the pixel is different, and the boundary pixels in the receptive field are enough to describe the modulation influence of the context on the central pixel, so that the radius parameter of the receptive field can not be adjusted, and the spatial action range of the receptive field can be reduced by reducing the radius parameter of the receptive field.
Step S504, enlarging the receptive field radius parameter according to a preset receptive field radius parameter enlarging rule.
In the embodiment of the present invention, the radial parameter of the receptive field is expanded according to a preset radial parameter expansion rule of the receptive field, and the spatial action range of the receptive field is generally expanded by expanding the radial parameter of the receptive field to k times.
As shown in fig. 6, a flowchart of a step of correcting a scale parameter provided in an embodiment of the present invention specifically includes the following steps:
step S602, determining whether the direction selection saliency is greater than a preset direction selection saliency threshold. When it is determined that the direction selection significance is greater than the preset direction selection significance threshold, performing step S604; when it is determined that the direction selection significance is not greater than the preset direction selection significance threshold, step S606 is performed.
And step S604, enlarging the scale parameters in the visual neuron energy response model according to a preset scale parameter enlargement rule.
And step S606, reducing the scale parameters in the visual neuron energy response model according to a preset scale parameter reduction rule.
In the embodiment of the invention, the fact that the direction selection significance is large indicates that the pixel is more likely to be a pixel for describing the outline characteristics, and the extraction effect of the outline characteristics by the coarse scale is better, so that for the pixel with the direction selection significance, the scale parameters in the visual neuron energy response model should be adaptively expanded, and the coarse scale visual neuron energy response model is utilized for processing, so as to improve the extraction effect of the outline characteristics of the pixel; on the contrary, the low direction selection significance indicates that the pixel is more likely to be a pixel for describing texture features, and the fine scale has a better inhibition effect on the texture features, so that for the pixel with the low direction selection significance, the scale parameters in the visual neuron energy response model should be adaptively reduced, and the fine scale visual neuron energy response model is utilized for processing, so as to improve the inhibition effect on the texture features of the pixel. In particular, it is considered that the preset visual neuron energy response model itself can be set as a coarse-scale or fine-scale visual neuron energy response model, and therefore, only one of the cases needs to be corrected in a specific case. For example, when the preset visual neuron energy response model has been processed by using a coarse-scale visual neuron energy response model, only step S606 needs to be executed; when the preset visual neuron energy response model is already processed by adopting the fine-scale visual neuron energy response model, only step S604 needs to be executed.
Fig. 7 is a schematic structural diagram of an image edge detection apparatus according to an embodiment of the present invention, which is described in detail below.
In an embodiment of the present invention, the image edge detection apparatus includes:
the energy response result calculating module 810 is configured to determine, according to a preset visual neuron energy response model, energy response results of pixel points in the image to be processed in a plurality of preset response orientations.
In the embodiment of the present invention, the preset visual neuron energy response model may be a visual neuron energy response model determined in advance through machine learning, wherein the visual neuron energy response model should include a scale parameter and also a response orientation parameter. Specifically, the energy response result of each pixel point in the image to be processed can be determined by performing convolution processing on the image to be processed by using the visual neuron energy response model, and when the response orientation parameter in the visual neuron energy response model is changed, the energy response result of each pixel point in the image to be processed in a plurality of preset response orientations can be obtained.
In a preferred embodiment of the present invention, considering that the Gabor filter and the LogGabor filter exhibit significant directional selectivity, preferably, the visual neuron energy response model is established in advance based on the Gabor filter or the LogGabor filter, and may also exhibit significant directional selectivity in a process of processing an image to be processed, so as to ensure that energy response results of each pixel point in different corresponding orientations have significant characteristics.
In the embodiment of the present invention, the visual neuron energy response model established in advance based on the Gabor filter is specifically as follows:
Figure BDA0002463819120000171
wherein, the meaning of a plurality of parameters contained in the visual neuron energy response model is specifically as follows: λ is the wavelength, 1/λ is the spatial frequency of the cosine function; sigma is the standard deviation of a Gaussian function of the size of the elliptical receptive field, and sigma/lambda represents the bandwidth of the spatial frequency, namely a scale parameter, wherein the larger the bandwidth of the spatial frequency is, the larger the scale of the visual neuron energy response model is, and the smaller the bandwidth of the spatial frequency is, the smaller the scale of the visual neuron energy response model is; theta is a neuron preference direction, namely the response orientation parameter, and the energy response results of pixel points in the image to be processed on a plurality of preset response orientations (theta) can be determined by adjusting theta; gamma is the ratio of the length to the length of the oval receptive field;
Figure BDA0002463819120000181
is a phase parameter;
Figure BDA0002463819120000182
Figure BDA0002463819120000183
at this time, the formula for performing convolution processing on the gray-scale image I (x, y) of the image to be processed according to the visual neuron energy response model is specifically as follows:
Figure BDA0002463819120000184
wherein, the mode of Gabor filter with phase parameter difference of pi/2 to image response is energy response result E for describing basic characteristics of each pixel pointλ,σ,θ,γ(x, y), specifically:
Figure BDA0002463819120000185
it can be seen that Eλ,σ,θ,γ(x, y) is
Figure BDA0002463819120000186
And
Figure BDA0002463819120000187
the two Gabor filters differ in phase parameter by pi/2 modulo the response of the image.
Further, let NθA preset response direction thetai(i.e., the neuron preference direction parameter θ of the visual neuron energy response model) is:
Figure BDA0002463819120000188
at the moment, energy response results of pixel points in the image to be processed on a plurality of preset response directions can be determined
Figure BDA0002463819120000189
In the embodiment of the present invention, the energy response model of the visual neuron established in advance based on the LogGabor filter is specifically as follows:
Figure BDA00024638191200001810
similarly, the meaning of the plurality of parameters included in the visual neuron energy response model is specifically as follows: f. of0Is the filter center frequency; sigmafFor radial bandwidth, σθThe directional bandwidth is a scale parameter, the larger the bandwidth of the radial bandwidth and the directional bandwidth is, the larger the scale of the visual neuron energy response model is, and the smaller the bandwidth of the radial bandwidth and the directional bandwidth is, the smaller the scale of the visual neuron energy response model is; theta0For the filter direction, i.e. the response orientation parameter mentioned above, the passOver-adjusting theta0And determining energy response results of pixel points in the image to be processed in a plurality of preset response directions. At the moment, convolution processing is carried out on the gray level image I (x, y) of the image to be processed according to the visual neuron energy response model, and an energy response result E describing the basic characteristics of each pixel point is obtainedf,θThe specific formula of (x, y) is:
Figure BDA0002463819120000191
hoddcorresponding to the real part in the H spatial domain of the filter, even filter, HevenThe imaginary part in the H spatial domain of the corresponding filter is an odd filter.
Further, N may be the sameθA preset response direction thetai(i.e., neuron preference direction parameter θ of visual neuron energy response model0) Comprises the following steps:
Figure BDA0002463819120000192
at the moment, energy response results of pixel points in the image to be processed on a plurality of preset response directions can be determined
Figure BDA0002463819120000193
In the embodiment of the present invention, it should be noted that, in the prior art, after determining the energy response results in the plurality of preset response orientations, the maximum energy response result is directly obtained
Figure BDA0002463819120000194
Or
Figure BDA0002463819120000195
As a result of the optimal energy response, corresponding response orientation
Figure BDA0002463819120000196
Or
Figure BDA0002463819120000197
And as an optimal corresponding direction, bringing the optimal energy response result and the optimal corresponding direction determined by the prior art into a non-classical receptive field edge detection algorithm to perform edge detection calculation.
Figure BDA0002463819120000198
Figure BDA0002463819120000199
Figure BDA0002463819120000201
Figure BDA0002463819120000202
And a direction selection significance calculation module 820, configured to determine the direction selection significance of the pixel point according to the energy response result and a preset direction selection significance calculation rule.
In the embodiment of the present invention, compared with the prior art in which the optimal energy response result and the optimal response orientation of each pixel point in the image are directly determined, the present invention further determines the direction selection saliency of the pixel point by using the energy response result and a preset direction selection saliency calculation rule, and for convenience of understanding, the foregoing visual neuron energy response model established based on the Gabor filter is used as an example to describe the calculation rule of the direction selection saliency of the pixel point. It should be noted that the calculation rule of the direction selection significance may also be extended to other optic neuron energy response models, such as an optic neuron energy response model established based on a LogGabor filter, which may be made by those skilled in the art according to common knowledge.
In the embodiment of the present invention, the direction selection saliency of a pixel point is determined by calculating the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions, and a specific calculation formula of the direction selection saliency of the pixel point is as follows:
Figure BDA0002463819120000203
the numerator on the right of the equation is the maximum energy response result of the pixel point in a plurality of preset directions, and the denominator is the average energy response result of the pixel point in the plurality of preset directions. It can be understood that the greater the direction selection significance, i.e., the greater the quotient of the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions, the more obvious the response of the pixel point in a certain direction is, and the more likely the pixel point is to be the contour feature of the image, and the smaller the direction selection significance, i.e., the greater the quotient of the maximum energy response result and the average energy response result of the pixel point in a plurality of preset directions is, the more likely the pixel point is to be the texture feature of the image.
And the scale parameter correction module 830 is configured to correct the scale parameter in the visual neuron energy response model according to the preset scale parameter correction rule and the size relationship between the direction selection significance and the preset direction selection significance threshold, and to generate a corrected visual neuron energy response model.
In the embodiment of the invention, the direction of each pixel point is selected to be the significance Sλ,σ(x, y) and a preset direction selection saliency threshold tsAnd comparing to determine whether the pixel point is a direction significant pixel point which has a particularly obvious response in a certain direction and is more likely to represent important structures such as a target contour and the like, or a non-direction significant pixel point which has a relatively balanced response in each direction and is more likely to represent non-important structures such as texture features and the like.
In the embodiment of the invention, considering that the image is processed by using the fine-scale (i.e. the scale parameter is relatively small) visual neuron energy response model to have a good inhibition effect on the texture, and the image is processed by using the coarse-scale (i.e. the scale parameter is relatively large) visual neuron energy response model to have a good extraction effect on the outline, therefore, after each pixel point is determined to be a direction significant pixel point or a non-direction significant pixel point, the scale parameter contained in the visual neuron energy response model is adaptively corrected, the non-direction significant pixel point is processed by using the fine-scale visual neuron energy response model, and the direction significant pixel point is processed by using the coarse-scale visual neuron energy response model.
In the embodiment of the present invention, it is emphasized that the coarse scale and the fine scale referred in the present invention only represent relative concepts, and the magnitude of the scale parameter is not particularly limited, the coarse scale visual neuron energy response model and the fine scale coarse scale visual neuron energy response model can be represented, but the scale parameter of the visual neuron energy response model is modified according to the directional significance of the pixel point, so that the technical scheme of respectively processing the pixel point with (without) directional significance by using the visual neuron energy response models with two or more different sizes of the scale parameter is within the scope claimed in the present invention.
And the optimal energy response result and orientation calculation module 840 is used for determining the optimal energy response result and the optimal response orientation of the pixel points according to the modified visual neuron energy response model.
In the embodiment of the present invention, each pixel point is processed based on the visual neuron energy response model after the scale parameter correction, and an optimal energy response result and an optimal response direction of the pixel point are further determined, where the calculation rules of the optimal energy response result and the optimal response direction are described in the foregoing step S102. That is to say, the scale parameters in the visual neuron energy response model processed for each pixel point are corrected based on the direction selection significance of the pixel point, so that a good suppression or extraction effect on each pixel point is ensured.
And an edge detection result determining module 850, configured to determine an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel point.
In the embodiment of the invention, a non-classical receptive field edge detection algorithm considering the influence of context modulation is utilized, and the optimal energy response result and the optimal response direction of each pixel point in the image are combined, so that the energy response result of each pixel point can be further corrected by utilizing the context, the corrected energy response result of each pixel point is displayed, and the edge detection result of the image to be processed can be obtained.
The image edge detection device provided by the embodiment of the invention determines the energy response results of pixel points in an image to be processed in a plurality of preset response orientations according to a visual neuron energy response model, further calculates the direction selection significance of the pixel points, judges whether the pixel points are direction significant pixel points which have significance in a certain orientation and are more likely to represent important structures such as a target contour or non-direction significant pixel points which are more likely to represent non-important structures such as texture features and the like in each orientation by using the magnitude relation between the direction selection significance and a preset direction selection significance threshold value, corrects the visual neuron energy response model according to different scale parameter correction rules aiming at the direction significant pixel points and the non-direction significant pixel points, and respectively calculates the optimal energy response result and the optimal response result of the corresponding pixel points by using the corrected visual neuron energy response model And finally determining an edge detection result by combining a non-classical receptive field edge detection algorithm. Compared with the prior art, the image edge detection device provided by the invention directly utilizes the same visual neuron energy response model to determine the optimal energy response result and the optimal response direction of each pixel point, the visual neuron energy response model is adjusted according to the direction selection significance of the pixel points, different visual neuron energy response models with scale parameters are used for the pixel points with different properties, the determined optimal energy response result and optimal response direction have better effect when being used for a non-classical receptive field edge detection algorithm, and the inhibition effect of textures in the edge detection result and the extraction effect of outlines are better.
FIG. 8 is a diagram illustrating an internal structure of a computer device in one embodiment. As shown in fig. 8, the computer apparatus includes a processor, a memory, a network interface, an input device, and a display screen connected through a system bus. Wherein the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program that, when executed by the processor, causes the processor to implement the image edge detection method. The internal memory may also have a computer program stored therein, which when executed by the processor, causes the processor to perform the image edge detection method. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, the image edge detection apparatus provided in the present application may be implemented in the form of a computer program, and the computer program may be run on a computer device as shown in fig. 8. The memory of the computer device may store various program modules constituting the image edge detection apparatus, such as an energy response result calculation module, a direction selection saliency calculation module, a scale parameter correction module, an optimal energy response result, an orientation calculation module, and an edge detection result determination module shown in fig. 7. The respective program modules constitute computer programs that cause the processors to execute the steps in the image edge detection methods of the respective embodiments of the present application described in the present specification.
For example, the computer device shown in fig. 8 may execute step S102 by the energy response result calculation module in the image edge detection apparatus shown in fig. 7; the computer device may perform step S104 by the direction selection saliency calculation module; the computer device may perform step S106 and so on through the scale parameter modification module.
In one embodiment, a computer device is proposed, the computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
determining energy response results of pixel points in the image to be processed in a plurality of preset response orientations according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
according to the magnitude relation between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, correcting the scale parameter in the visual neuron energy response model to generate a corrected visual neuron energy response model;
determining the optimal energy response result and the optimal response direction of the pixel points according to the corrected visual neuron energy response model;
and determining an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the steps of:
determining energy response results of pixel points in the image to be processed in a plurality of preset response orientations according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
according to the magnitude relation between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, correcting the scale parameter in the visual neuron energy response model to generate a corrected visual neuron energy response model;
determining the optimal energy response result and the optimal response direction of the pixel points according to the corrected visual neuron energy response model;
and determining an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
It should be understood that, although the steps in the flowcharts of the embodiments of the present invention are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in various embodiments may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a non-volatile computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the program is executed. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims (10)

1. An image edge detection method, comprising:
determining energy response results of pixel points in the image to be processed in a plurality of preset response orientations according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
according to the magnitude relation between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, correcting the scale parameter in the visual neuron energy response model to generate a corrected visual neuron energy response model;
determining the optimal energy response result and the optimal response direction of the pixel points according to the corrected visual neuron energy response model;
and determining an edge detection result of the image to be processed according to a non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
2. The image edge detection method according to claim 1, wherein the step of determining the edge detection result of the image to be processed according to the non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response orientation of the pixel point specifically comprises:
determining boundary pixel points according to the determined central pixel points and preset receptive field radius parameters; the distance between the boundary pixel point and the central pixel point is smaller than the receptive field radius parameter;
determining an offset angle and a distance between the central pixel point and the boundary pixel point; the offset angle refers to an angle difference between the optimal response direction of the central pixel point and the optimal response direction of the boundary pixel point;
determining an energy response influence value of the boundary pixel point on the central pixel point according to the offset angle and the distance and an optimal energy response result of the boundary pixel point;
determining the complete energy response of the central pixel point according to the energy response influence value and the optimal energy response of the central pixel point;
and determining an edge detection result of the image to be processed according to the complete energy response.
3. The image edge detection method according to claim 2, wherein the step of determining boundary pixel points according to the determined central pixel points and preset receptive field radius parameters specifically comprises:
determining the local contrast of the central pixel point according to the gray level image of the image to be processed;
correcting the receptive field radius parameter according to the magnitude relation between the local contrast and a preset low-contrast response threshold and a preset receptive field radius parameter correction rule to generate a corrected receptive field radius parameter;
and determining boundary pixel points according to the central pixel points and the corrected receptive field radius parameters.
4. The image edge detection method according to claim 3, wherein the step of determining the local contrast of the central pixel point according to the gray-scale image of the image to be processed specifically comprises:
determining a gray level image within the receptive field range of the central pixel point according to a preset receptive field radius parameter;
determining the maximum brightness value and the minimum brightness value of the gray level image in the receptive field range;
and determining the local contrast of the central pixel point according to the maximum brightness value and the minimum brightness value.
5. The image edge detection method according to claim 3, wherein the step of correcting the receptive field radius parameter according to the magnitude relationship between the local contrast and a preset low-contrast response threshold and a preset receptive field radius parameter correction rule specifically comprises:
judging whether the local contrast is smaller than a preset low-contrast response threshold value or not;
and when the local contrast is judged to be smaller than a preset low-contrast response threshold, enlarging the receptive field radius parameter according to a preset receptive field radius parameter enlargement rule.
6. The image edge detection method according to claim 1, wherein the step of determining, according to a preset visual neuron energy response model, energy response results of pixel points in the image to be processed in a plurality of preset response orientations specifically includes:
and determining energy response results of pixel points in the image to be processed in a plurality of preset response directions according to a visual neuron energy response model established in advance based on a Gabor filter or a LogGabor filter.
7. The image edge detection method according to claim 1, wherein the step of correcting the scale parameter in the visual neuron energy response model according to the magnitude relationship between the direction-selected saliency and the preset direction-selected saliency threshold and a preset scale parameter correction rule specifically includes:
judging whether the direction selection significance is larger than a preset direction selection significance threshold value or not;
when the direction selection significance is judged to be larger than a preset direction selection significance threshold value, expanding scale parameters in the visual neuron energy response model according to a preset scale parameter expansion rule; or
When the direction selection significance is judged to be not greater than a preset direction selection significance threshold value, reducing scale parameters in the visual neuron energy response model according to a preset scale parameter reduction rule; or
When the direction selection significance is judged to be larger than a preset direction selection significance threshold value, expanding scale parameters in the visual neuron energy response model according to a preset scale parameter expansion rule;
and when the direction selection significance is judged to be not greater than a preset direction selection significance threshold value, reducing the scale parameters in the visual neuron energy response model according to a preset scale parameter reduction rule.
8. An image edge detection apparatus, comprising:
the energy response result calculation module is used for determining energy response results of pixel points in the image to be processed in a plurality of preset response directions according to a preset visual neuron energy response model; the visual neuron energy response model comprises scale parameters;
the direction selection significance calculation module is used for determining the direction selection significance of the pixel points according to the energy response result and a preset direction selection significance calculation rule;
the scale parameter correction module is used for correcting the scale parameter in the visual neuron energy response model according to the size relationship between the direction selection significance and a preset direction selection significance threshold value and a preset scale parameter correction rule, and generating a corrected visual neuron energy response model;
the optimal energy response result and orientation calculation module is used for determining the optimal energy response result and the optimal response orientation of the pixel points according to the corrected visual neuron energy response model;
and the edge detection result determining module is used for determining the edge detection result of the image to be processed according to the non-classical receptive field edge detection algorithm and the optimal energy response result and the optimal response direction of the pixel points.
9. A computer device comprising a memory and a processor, the memory having stored therein a computer program which, when executed by the processor, causes the processor to carry out the steps of the image edge detection method according to any one of claims 1 to 7.
10. A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor, causes the processor to carry out the steps of the image edge detection method according to any one of claims 1 to 7.
CN202010327742.0A 2020-04-23 2020-04-23 Image edge detection method, device, computer equipment and storage medium Active CN111539969B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010327742.0A CN111539969B (en) 2020-04-23 2020-04-23 Image edge detection method, device, computer equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010327742.0A CN111539969B (en) 2020-04-23 2020-04-23 Image edge detection method, device, computer equipment and storage medium

Publications (2)

Publication Number Publication Date
CN111539969A true CN111539969A (en) 2020-08-14
CN111539969B CN111539969B (en) 2023-06-09

Family

ID=71978931

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010327742.0A Active CN111539969B (en) 2020-04-23 2020-04-23 Image edge detection method, device, computer equipment and storage medium

Country Status (1)

Country Link
CN (1) CN111539969B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112486636A (en) * 2020-12-28 2021-03-12 成都辰迈科技有限公司 Cloud computing system for cloud computing network management
CN117495851A (en) * 2023-12-29 2024-02-02 陕西中医药大学 Image contour processing-based water environment microorganism detection method

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030081215A1 (en) * 2001-01-09 2003-05-01 Ajay Kumar Defect detection system for quality assurance using automated visual inspection
CN103903251A (en) * 2012-12-30 2014-07-02 南京理工大学 Night vision image salient contour extracting method based on non-classical receptive field composite modulation
EP2942753A1 (en) * 2014-05-05 2015-11-11 Xiaomi Inc. Method and device for image segmentation
CN106033610A (en) * 2016-03-22 2016-10-19 广西科技大学 Contour detection method based on non-classical receptive field space summation modulation
CN107767387A (en) * 2017-11-09 2018-03-06 广西科技大学 Profile testing method based on the global modulation of changeable reception field yardstick
CN107909593A (en) * 2017-11-07 2018-04-13 广西科技大学 Non- set direction profile testing method based on receptive field subregion
CN108053415A (en) * 2017-12-14 2018-05-18 广西科技大学 Based on the bionical profile testing method for improving non-classical receptive field
CN108090492A (en) * 2017-11-09 2018-05-29 广西科技大学 The profile testing method inhibited based on scale clue
EP3413232A1 (en) * 2017-06-09 2018-12-12 Ricoh Company Ltd. Method, device, and apparatus for detecting road dividing object as well as computer program and non-transitory computer-readable medium
CN110210493A (en) * 2019-04-30 2019-09-06 中南民族大学 Profile testing method and system based on non-classical receptive field modulation neural network

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030081215A1 (en) * 2001-01-09 2003-05-01 Ajay Kumar Defect detection system for quality assurance using automated visual inspection
CN103903251A (en) * 2012-12-30 2014-07-02 南京理工大学 Night vision image salient contour extracting method based on non-classical receptive field composite modulation
EP2942753A1 (en) * 2014-05-05 2015-11-11 Xiaomi Inc. Method and device for image segmentation
CN106033610A (en) * 2016-03-22 2016-10-19 广西科技大学 Contour detection method based on non-classical receptive field space summation modulation
EP3413232A1 (en) * 2017-06-09 2018-12-12 Ricoh Company Ltd. Method, device, and apparatus for detecting road dividing object as well as computer program and non-transitory computer-readable medium
CN107909593A (en) * 2017-11-07 2018-04-13 广西科技大学 Non- set direction profile testing method based on receptive field subregion
CN107767387A (en) * 2017-11-09 2018-03-06 广西科技大学 Profile testing method based on the global modulation of changeable reception field yardstick
CN108090492A (en) * 2017-11-09 2018-05-29 广西科技大学 The profile testing method inhibited based on scale clue
CN108053415A (en) * 2017-12-14 2018-05-18 广西科技大学 Based on the bionical profile testing method for improving non-classical receptive field
CN110210493A (en) * 2019-04-30 2019-09-06 中南民族大学 Profile testing method and system based on non-classical receptive field modulation neural network

Non-Patent Citations (9)

* Cited by examiner, † Cited by third party
Title
GRIGORESCU C等: "Contour detection based on nonclassical receptive field inhibition", IEEE TRANSACTIONS ON IMAGE PROCESSING *
LI LONG等: "Contour detection based on the property of orientation selective inhibition of non- classical receptive field", 2008 IEEE CONFERENCE ON CYBERNETICS AND INTELLIGENT SYSTEMS *
于江波;陈后金;李居朋;: "视觉选择注意机制模型及其应用", 电子测量与仪器学报, no. 04 *
朱文杰;王广龙;高凤岐;乔中涛;黄瑞;: "引导滤波与视觉感知结合的自然场景轮廓提取", 系统工程与电子技术, no. 01 *
林川;李亚;曹以隽;: "考虑微动机制与感受野特性的轮廓检测模型", 计算机工程与应用, no. 24 *
武薇;周涛;朱亚萍;范影乐;: "引入初级视通路计算模型的轮廓检测", 中国图象图形学报, no. 05 *
肖洁;蔡超;郭照立;: "空间统一调制模型的轮廓检测" *
肖洁;蔡超;郭照立;: "空间统一调制模型的轮廓检测", 中国图象图形学报, no. 11 *
蒋涯;武薇;范影乐;: "结合对比度适应与侧抑制信息关联的轮廓检测方法", 航天医学与医学工程, no. 05 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112486636A (en) * 2020-12-28 2021-03-12 成都辰迈科技有限公司 Cloud computing system for cloud computing network management
CN117495851A (en) * 2023-12-29 2024-02-02 陕西中医药大学 Image contour processing-based water environment microorganism detection method
CN117495851B (en) * 2023-12-29 2024-04-05 陕西中医药大学 Image contour processing-based water environment microorganism detection method

Also Published As

Publication number Publication date
CN111539969B (en) 2023-06-09

Similar Documents

Publication Publication Date Title
CN108805828B (en) Image processing method, device, computer equipment and storage medium
JP7030493B2 (en) Image processing equipment, image processing methods and programs
CN109658344A (en) Image de-noising method, device, equipment and storage medium based on deep learning
CN111415317B (en) Image processing method and device, electronic equipment and computer readable storage medium
CN111539969A (en) Image edge detection method and device, computer equipment and storage medium
CN112150371B (en) Image noise reduction method, device, equipment and storage medium
CN113674191A (en) Weak light image enhancement method and device based on conditional countermeasure network
CN112950497A (en) Image processing method, image processing device, electronic equipment and storage medium
CN111681165A (en) Image processing method, image processing device, computer equipment and computer readable storage medium
CN108229650B (en) Convolution processing method and device and electronic equipment
CN111860582A (en) Image classification model construction method and device, computer equipment and storage medium
CN114820363A (en) Image processing method and device
CN109447935B (en) Infrared image processing method and device, computer equipment and readable storage medium
CN111340025A (en) Character recognition method, character recognition device, computer equipment and computer-readable storage medium
US20170140538A1 (en) Image preprocessing method and electronic device for image registration
CN113936163B (en) Image processing method, terminal and storage medium
CN109447942A (en) Image blur determines method, apparatus, computer equipment and storage medium
CN115861044A (en) Complex cloud layer background simulation method, device and equipment based on generation countermeasure network
US20070081193A1 (en) Method and apparatus for expanding bit resolution using local information of image
CN111881907B (en) Frame regression positioning method and device and electronic equipment
CN115147296A (en) Hyperspectral image correction method, device, computer equipment and storage medium
CN115797194A (en) Image denoising method, image denoising device, electronic device, storage medium, and program product
CN112200730B (en) Image filtering processing method, device, equipment and storage medium
CN114399495A (en) Image definition calculation method, device, equipment and storage medium
CN111914779A (en) Table text detection method and device, computer equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant