Microscopic image optimization method based on biological visual characteristics

Abstract

The invention discloses a microscopic image optimization method based on biological visual characteristics, and relates to the technical field of microbiological microscopic imaging, and the method comprises the following steps: S1, according to the characteristics of an input image, adaptively adjusting the parameters of a retina model; s2, performing multi-level central visual channel and peripheral visual channel processing on the input image by using a retina model; and S3, fusing the output data of the central visual channel and the peripheral visual channel to generate an optimized image. According to the invention, the working principle of a human visual system, especially the space-time filtering characteristic of the retina, is simulated, the efficient optimization processing of the microbiological microscopic image is realized, the problems of low contrast ratio, incomplete noise suppression and difficult overlapping region separation of the existing microbiological microscopic image are solved, and the high-fidelity imaging of dense microbiological populations is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of microbial microscopic imaging, and in particular to a microscopic image optimization method based on biological visual characteristics. Background Art

[0002] Microbial microscopy technology has widespread applications in the biomedical field. With the deepening integration of sciences like artificial intelligence and biomedicine, the demand for microscopic image quality is increasing. High-quality microbial microscopy images can provide more accurate sample characteristics for AI deep learning models. However, due to the influence of factors such as optical system noise and the characteristics of the sample itself during the microscopic imaging process, the raw microscopic images obtained often suffer from low contrast, high noise, and blurred edges, requiring optimization.

[0003] Traditional microscopic image optimization methods based on histogram equalization and Gaussian filtering have significant drawbacks: histogram equalization tends to amplify noise while enhancing contrast, linear processing leads to unnatural dynamic range expansion, overexposure in bright areas such as the center of microbial aggregates, and loss of detail in dim areas such as the edges of individual microorganisms. Gaussian filtering also blurs image edges while removing noise, resulting in loss of detail and failing to meet the requirements for high-quality microscopic images. Summary of the Invention

[0004] In response to the above-mentioned deficiencies in the prior art, the microscopic image optimization method based on biological visual characteristics provided by the present invention solves the problems that traditional microscopic image optimization cannot effectively optimize microbial microscopic images, has low contrast, incomplete noise suppression, and difficulty in separating overlapping areas.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is: The microscopic image optimization method based on biological visual characteristics includes the following steps: S1, adaptively adjust the parameters of the retinal model according to the characteristics of the input image; S2, using the retinal model to process the input image in multiple levels of central and peripheral visual channels; S3. Fusing the output data of the central visual channel and the peripheral visual channel to generate an optimized image.

[0006] Furthermore, the S1 includes: The local adaptation sensitivity of the photoreceptors used to control the strength of the local brightness adaptation of the image is adjusted by the following formula: , Where pLAS is the local adaptation sensitivity of the photoreceptor, min(,) is the function to find the minimum value, σ Lis the brightness standard deviation of the input image, μ L is the mean brightness of the input image; The horizontal cell gain, which is used to control the intensity of the image halo effect and thus affect the edge enhancement effect, is adjusted by the following formula: , Among them, hCG is the horizontal cell gain, max(,) is the function to find the maximum value, C local is the local contrast of the input image; The bipolar cell contrast enhancement degree used to control the image contrast enhancement effect is adjusted by the following formula: , Among them, bCC is the bipolar cell contrast enhancement; The sensitivity of the central visual channel of ganglion cells used to control image detail enhancement is adjusted by the following formula: , Among them, cVGCS is the sensitivity of the central visual channel of ganglion cells, N level is the noise level of the input image calculated using wavelet transform; The sensitivity of the peripheral visual channel of ganglion cells used to control image motion detection is adjusted by the following formula: , Among them, pVGCS is the sensitivity of the peripheral visual channel of ganglion cells, D frame Evaluate the temporal variation of the input image.

[0007] Furthermore, the method for calculating the local contrast of the input image includes the following steps: A1. For each pixel in the input image, a local window is defined that surrounds the pixel at its center. A2. Calculate the standard deviation and mean of all pixels in the local window, and divide the standard deviation by the mean as the contrast of the local window; A3. averaging the contrasts of the local windows corresponding to all pixels of the input image to obtain the local contrast of the input image.

[0008] Furthermore, the retinal models of S1 and S2 include: The photoreceptor layer is used to simulate the cones and rods in the biological retina and perform local brightness adaptation on the input image; The horizontal cell layer is used to simulate the horizontal cells in the biological retina, providing lateral inhibition signals and performing local contrast enhancement on the output data of the photoreceptor layer; The bipolar cell layer is used to simulate the bipolar cells in the biological retina, and further integrates the output data of the horizontal cell layer and enhances the contrast.

[0009] The ganglion cell layer, used to simulate the ganglion cells in the biological retina, processes the output data of the bipolar cell layer through the central visual channel and the peripheral visual channel.

[0010] Furthermore, the operation process expression of the photoreceptor layer is: , Among them, P adapted (x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t after local brightness adaptive adjustment, ε1 is the first constant used to prevent division by zero, pLAS is the local adaptation sensitivity of the photoreceptor, and P(x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t without local brightness adaptive adjustment; , Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, * is the convolution operation, I denoised (x,y,t) is the pixel value at the (x,y) coordinate of the input image at time t.

[0011] Furthermore, the operation process expression of the horizontal cell layer is: , Among them, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, P adapted (x, y, t) is the (x, y) coordinate pixel output value of the photoreceptor layer after local brightness adaptive adjustment at time t, hCG is the horizontal cell gain, and H(x, y, t) is the (x, y) coordinate value of the lateral inhibition signal at time t; , Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, and * is the convolution operation.

[0012] Furthermore, the operational expression of the bipolar cell layer is: , Among them, IPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, B ON_enhanced(x, y, t) is the pixel contrast enhancement value of the (x, y) coordinate of the ON polar cell at time t, B OFF_enhanced (x, y, t) is the (x, y) coordinate pixel contrast enhancement value of the OFF polarity cell at time t; The ON-type polarity cells are used to respond to increased image brightness; The OFF-type polarity cells are used to respond to image brightness reduction; , , Among them, B ON (x, y, t) is the (x, y) coordinate value of ON polar cell at time t, B OFF (x, y, t) is the (x, y) coordinate value of the OFF polar cell at time t, and hCG is the horizontal cell gain; , , Where max(,) is the function for finding the maximum value, and OPL(x,y,t) is the pixel output value of the (x,y) coordinate of the horizontal cell layer at time t.

[0013] Furthermore, the operation process expression of the central visual channel of the ganglion cell layer is: , Where CV (x, y, t) is the (x, y) coordinate pixel output value of the central visual channel of the ganglion cell layer at time t, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, cVGCS is the sensitivity of the central visual channel of the ganglion cell, and ε2 is a second constant used to prevent division by zero; The computational expression of the peripheral visual channel of the ganglion cell layer is: , Where PV (x, y, t) is the (x, y) coordinate pixel output value of the peripheral visual channel of the ganglion cell layer at time t, ε3 is the third constant used to prevent division by zero, and pVGCS is the sensitivity of the ganglion cell peripheral visual channel. is the operation of finding the partial derivative with respect to time.

[0014] Furthermore, the step S3 includes the following sub-steps: S31, converting the output data of the central visual channel from the BGR color space to the YCrCb color space; S32. Linearly map the output data of the peripheral visual channel of the ganglion cell layer to a value range of 0 to 255 using the following formula, so that it is aligned with the value range of the Y component of the YCrCb color space: , Among them, P norm is the image data after linear mapping, P is the image data before linear mapping, max() is the function for finding the maximum value, and min() is the function for finding the minimum value; S33, retaining the Cr component and Cb component of the output data of the central visual channel of the ganglion cell layer after conversion to the YCrCb color space, and using the output data of the peripheral visual channel of the ganglion cell layer after linear mapping as a new Y component, and fusing them to obtain an optimized image in the YCrCb space; S34: Convert the optimized image in the YCrCb space back to the BGR space to obtain the final optimized image.

[0015] The beneficial effects of the present invention are: (1) The present invention simulates the working principle of the human visual system, especially the spatiotemporal filtering characteristics of the retina, to achieve efficient optimization processing of microbial microscopic images. It solves the problems of low contrast, incomplete noise suppression, and difficulty in separating overlapping areas in existing microbial microscopic images, and realizes high-fidelity imaging of dense microbial populations.

[0016] (2) The present invention adjusts the parameters of the retinal model based on the features of the input image, ensuring that local adaptability is enhanced in images with large brightness changes, edges are enhanced in images with low contrast, the sensitivity of the central visual channel is increased in images with large noise, and the sensitivity of the peripheral visual channel is enhanced in sequences with obvious motion.

[0017] (3) The retinal model photoreceptor layer of the present invention simulates the local adaptation mechanism of cones and rods in the retina of an organism to achieve adaptive brightness adjustment in different areas of the image, so that the microbial features in overly bright and dark areas are clearly visible at the same time.

[0018] (4) The photoreceptor layer and horizontal cell layer of the retinal model of the present invention simulate the spectral whitening principle of the outer plexiform layer of the retina of biological organisms. Through spatiotemporal filtering, they reduce the low-frequency brightness energy, i.e., the average brightness, and enhance the mid-frequency brightness, i.e., the details, thereby effectively enhancing the detail information in the microbial microscopic image while retaining the morphological characteristics of the microorganisms.

[0019] (5) The photoreceptor layer and horizontal cell layer of the retinal model of the present invention are also based on the non-separable spatiotemporal filtering characteristics of the retina, effectively removing high-frequency spatial and temporal noise while retaining the edge information of microorganisms.

[0020] (6) The bipolar cell layer of the retinal model of the present invention improves the edge clarity of the overlapping area by simulating the contrast enhancement mechanism of the retina, which facilitates subsequent applications of microbial microscopic images such as microbial segmentation and counting.

[0021] (7) The ganglion cell layer of the retinal model of the present invention is responsible for detailed color vision and texture recognition through the central visual channel, corresponding to the small cell pathway of the biological retina; and is responsible for sensitive transient signal detection such as motion and event edge structure recognition through the peripheral visual channel, corresponding to the large cell pathway of the biological retina.

[0022] (8) The present invention fuses the output data of the central visual channel and the peripheral visual channel, retaining the Cr component and Cb component of the output data of the central visual channel of the ganglion cell layer after conversion to the YCrCb color space, that is, retaining the chromaticity information of the central visual channel; at the same time, the output data of the peripheral visual channel of the ganglion cell layer after linear mapping is used as the new Y component, that is, the motion and edge features of the peripheral visual channel are injected, making full use of the biological mechanism that the large cell pathway of the biological retina dominates motion perception and the small cell pathway is responsible for color recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of a microscopic image optimization method based on biological visual characteristics provided by an embodiment of the present invention; Figure 2 A schematic diagram of the biological retina structure used as a reference when setting the retina model in an embodiment of the present invention; Figure 3 A structural diagram of a retinal model according to an embodiment of the present invention; Figure 4 is an input image according to an embodiment of the present invention; Figure 5 The image output by the central visual channel of an embodiment of the present invention; Figure 6 An image output by the peripheral vision channel of an embodiment of the present invention; Figure 7 This is the final optimized image of the embodiment of the present invention. DETAILED DESCRIPTION

[0024] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0025] like Figure 1 As shown, in one embodiment of the present invention, a microscopic image optimization method based on biological visual characteristics includes the following steps: S1. Adaptively adjust the parameters of the retinal model based on the characteristics of the input image, including: The local adaptation sensitivity of the photoreceptors used to control the strength of the local brightness adaptation of the image is adjusted by the following formula: , Where pLAS is the local adaptation sensitivity of the photoreceptor, min(,) is the function to find the minimum value, σ L is the brightness standard deviation of the input image, μ L is the mean brightness of the input image; The horizontal cell gain, which is used to control the intensity of the image halo effect and thus affect the edge enhancement effect, is adjusted by the following formula: , Where hCG is the horizontal cell gain, max(,) is the function to find the maximum value; C local is the local contrast C of the input image local The calculation method comprises the following steps: A1, for each pixel in the input image, defining a local window enclosing the pixel at the exact center; A2, calculating the standard deviation and mean of all pixels in the local window, dividing the standard deviation by the mean as the contrast of the local window; A3, averaging the contrast of the local windows corresponding to all pixels of the input image to obtain the local contrast of the input image; The bipolar cell contrast enhancement degree used to control the image contrast enhancement effect is adjusted by the following formula: , Among them, bCC is the bipolar cell contrast enhancement; The sensitivity of the central visual channel of ganglion cells used to control image detail enhancement is adjusted by the following formula: , Among them, cVGCS is the sensitivity of the central visual channel of ganglion cells, N level is the noise level of the input image calculated using wavelet transform; The sensitivity of the peripheral visual channel of ganglion cells used to control image motion detection is adjusted by the following formula: , Among them, pVGCS is the sensitivity of the peripheral visual channel of ganglion cells, D frame Evaluate the temporal variation of the input image; If the input image is a dynamic video sequence, D is obtained by calculating the difference between frames. frame ; If the input image is a static picture, D frame is 0.

[0026] The present invention adjusts the retinal model parameters by using the features of the input image to ensure enhanced local adaptability in images with large brightness changes, enhanced edges in images with low contrast, increased sensitivity of the central visual channel in images with high noise, and enhanced sensitivity of the peripheral visual channel in sequences with obvious motion.

[0027] S2. Use the retinal model to process the input image at multiple levels of central and peripheral visual channels.

[0028] The human retina is an important component of the visual system, responsible for converting light signals into neural signals and performing preliminary processing.

[0029] like Figure 2 As shown in Figure 1, the biological retina consists of multiple layers of cells, primarily photoreceptors (cones and rods), horizontal cells, bipolar cells, amacrine cells, and ganglion cells. These cells form a complex neural network that performs a range of signal processing functions.

[0030] Photoreceptor cells synapse with horizontal and bipolar cells in the first synaptic layer, the outer plexiform layer (OPL), which contributes to contrast enhancement and color texture. In the second synaptic layer, the inner plexiform layer (IPL), bipolar and amacrine cells synapse with ganglion cells. Amacrine cells in the IPL are able to respond to moving objects.

[0031] Information processing in the biological retina is mainly carried out through two pathways: the central vision pathway, which is responsible for detailed color vision and texture recognition, and corresponds to the small cell pathway; the peripheral vision pathway, which is responsible for sensitive transient signal detection (such as motion and events) and edge structure recognition, and corresponds to the large cell pathway.

[0032] The main processing characteristics of the biological retina include: 1) local logarithmic brightness compression: implemented at the photoreceptor and ganglion cell level, so that both very bright and very dark areas can appear in clear detail in the same image; 2) spectral brightening of the outer plexiform layer: achieved through spatiotemporal filtering of photoreceptors and horizontal cells, reducing low-frequency brightness energy (average brightness) and enhancing mid-frequency brightness (detail); 3) high-frequency noise filtering: through non-separable spatiotemporal filtering, effectively reducing high-frequency spatial and temporal noise.

[0033] Based on this, Figure 3 As shown, the retinal model of the present invention includes: The photoreceptor layer is used to simulate the cones and rods in the biological retina and perform local brightness adaptation on the input image; The horizontal cell layer is used to simulate the horizontal cells in the biological retina, providing lateral inhibition signals and performing local contrast enhancement on the output data of the photoreceptor layer; The bipolar cell layer is used to simulate the bipolar cells in the biological retina, and further integrates the output data of the horizontal cell layer and enhances the contrast.

[0034] The ganglion cell layer, used to simulate the ganglion cells in the biological retina, processes the output data of the bipolar cell layer through the central visual channel and the peripheral visual channel.

[0035] The operation process expression of the photoreceptor layer is: , Among them, P adapted (x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t after local brightness adaptive adjustment, ε1 is the first constant used to prevent division by zero, pLAS is the local adaptation sensitivity of the photoreceptor, and P(x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t without local brightness adaptive adjustment; , Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, * is the convolution operation, I denoised (x,y,t) is the pixel value at the (x,y) coordinate of the input image at time t.

[0036] The retinal model photoreceptor layer of the present invention simulates the local adaptation mechanism of cones and rods in the retina of an organism to achieve adaptive brightness adjustment in different areas of the image, so that the microbial features in overly bright and dark areas are clearly visible at the same time.

[0037] The operation process expression of the horizontal cell layer is: , Among them, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, P adapted (x, y, t) is the (x, y) coordinate pixel output value of the photoreceptor layer after local brightness adaptive adjustment at time t, hCG is the horizontal cell gain, and H(x, y, t) is the (x, y) coordinate value of the lateral inhibition signal at time t; , Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, and * is the convolution operation.

[0038] The photoreceptor layer and horizontal cell layer of the retinal model of the present invention simulate the spectral whitening principle of the outer plexiform layer of the retina of biological organisms. Through spatiotemporal filtering, they reduce the low-frequency brightness energy, i.e., the average brightness, and enhance the mid-frequency brightness, i.e., the details, thereby effectively enhancing the detail information in the microbial microscopic image while retaining the morphological characteristics of the microorganisms.

[0039] The photoreceptor layer and horizontal cell layer are also based on the non-separable spatiotemporal filtering characteristics of the retina, effectively removing high-frequency spatial and temporal noise while retaining the edge information of microorganisms.

[0040] The operational expression of the bipolar cell layer is: , Among them, IPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, B ON_enhanced (x, y, t) is the pixel contrast enhancement value of the (x, y) coordinate of the ON polar cell at time t, B OFF_enhanced (x, y, t) is the (x, y) coordinate pixel contrast enhancement value of the OFF polarity cell at time t; ON-type polarity cells are used to respond to increases in image brightness; OFF-type polarity cells are used to respond to image brightness reduction; , , Among them, B ON (x, y, t) is the (x, y) coordinate value of ON polar cell at time t, B OFF (x, y, t) is the (x, y) coordinate value of the OFF polar cell at time t, and hCG is the horizontal cell gain; , , Where max(,) is the function for finding the maximum value, and OPL(x,y,t) is the pixel output value of the (x,y) coordinate of the horizontal cell layer at time t.

[0041] The retinal model bipolar cell layer of the present invention improves edge clarity of overlapping areas by simulating the contrast enhancement mechanism of the retina, thereby facilitating subsequent applications of microbial microscopic images such as microbial segmentation and counting.

[0042] The operation process expression of the central visual channel of the ganglion cell layer is: , Where CV (x, y, t) is the (x, y) coordinate pixel output value of the central visual channel of the ganglion cell layer at time t, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, cVGCS is the sensitivity of the central visual channel of the ganglion cell, and ε2 is a second constant used to prevent division by zero; The computational expression of the peripheral visual channel in the ganglion cell layer is: , Where PV (x, y, t) is the (x, y) coordinate pixel output value of the peripheral visual channel of the ganglion cell layer at time t, ε3 is the third constant used to prevent division by zero, and pVGCS is the sensitivity of the ganglion cell peripheral visual channel. is the operation of finding the partial derivative with respect to time.

[0043] The ganglion cell layer of the retinal model of the present invention is responsible for detailed color vision and texture recognition through the central visual channel, corresponding to the biological retinal small cell pathway; and is responsible for sensitive transient signal detection such as motion and event edge structure recognition through the peripheral visual channel, corresponding to the biological retinal large cell pathway.

[0044] S3, fusing the output data of the central visual channel and the peripheral visual channel to generate an optimized image, including the following steps: S31, converting the output data of the central visual channel from the BGR color space to the YCrCb color space; S32. Linearly map the output data of the peripheral visual channel of the ganglion cell layer to a value range of 0 to 255 using the following formula, so that it is aligned with the value range of the Y component of the YCrCb color space: , Among them, P norm is the image data after linear mapping, P is the image data before linear mapping, max() is the function for finding the maximum value, and min() is the function for finding the minimum value; S33, retaining the Cr component and Cb component of the output data of the central visual channel of the ganglion cell layer after conversion to the YCrCb color space, and using the output data of the peripheral visual channel of the ganglion cell layer after linear mapping as a new Y component, and fusing them to obtain an optimized image in the YCrCb space; S34: Convert the optimized image in the YCrCb space back to the BGR space to obtain the final optimized image.

[0045] The invention fuses the output data of the central visual channel and the peripheral visual channel, retaining the Cr component and Cb component of the central visual channel output data of the ganglion cell layer after conversion to the YCrCb color space, that is, retaining the chromaticity information of the central visual channel; at the same time, the output data of the peripheral visual channel of the ganglion cell layer after linear mapping is used as the new Y component, that is, the motion and edge features of the peripheral visual channel are injected, making full use of the biological mechanism that the large cell pathway of the biological retina dominates motion perception and the small cell pathway is responsible for color recognition.

[0046] In this embodiment, the input image to be processed is as follows: Figure 4 As shown, the image output by the central visual channel is as follows Figure 5 As shown, the image output by the peripheral vision channel is as follows Figure 6 As shown in the figure, the final optimized image is as follows Figure 7 shown.

[0047] In summary, the present invention simulates the working principles of the human visual system, especially the spatiotemporal filtering characteristics of the retina, to achieve efficient optimization processing of microbial microscopic images. It solves the problems of low contrast, incomplete noise suppression, and difficulty in separating overlapping areas in existing microbial microscopic images, and realizes high-fidelity imaging of dense microbial populations.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A microscopic image optimization method based on biological visual characteristics, characterized in that: The following steps are involved: S1, adaptively adjust the parameters of the retinal model according to the characteristics of the input image; S2, using the retinal model to process the input image in multiple levels of central and peripheral visual channels; S3. Fusing the output data of the central visual channel and the peripheral visual channel to generate an optimized image.

2. The microscopic image optimization method based on biological visual characteristics according to claim 1, characterized in that: Said S1 comprises: The local adaptation sensitivity of the photoreceptors used to control the strength of the local brightness adaptation of the image is adjusted by the following formula: ， Where pLAS is the local adaptation sensitivity of the photoreceptor, min(,) is the function to find the minimum value, σ L is the brightness standard deviation of the input image, μ L is the mean brightness of the input image; The horizontal cell gain, which is used to control the intensity of the image halo effect and thus affect the edge enhancement effect, is adjusted by the following formula: ， Among them, hCG is the horizontal cell gain, max(,) is the function to find the maximum value, C local is the local contrast of the input image; The bipolar cell contrast enhancement degree used to control the image contrast enhancement effect is adjusted by the following formula: ， Among them, bCC is the bipolar cell contrast enhancement; The sensitivity of the central visual channel of ganglion cells used to control image detail enhancement is adjusted by the following formula: ， Among them, cVGCS is the sensitivity of the central visual channel of ganglion cells, N level is the noise level of the input image calculated using wavelet transform; The sensitivity of the peripheral visual channel of ganglion cells used to control image motion detection is adjusted by the following formula: ， Among them, pVGCS is the sensitivity of the peripheral visual channel of ganglion cells, D frame Evaluate the temporal variation of the input image.

3. The microscopic image optimization method based on biological visual characteristics according to claim 2, characterized in that: The method for calculating the local contrast of the input image comprises the following steps: A1. For each pixel in the input image, a local window is defined that surrounds the pixel at its center. A2. Calculate the standard deviation and mean of all pixels in the local window, and divide the standard deviation by the mean as the contrast of the local window; A3. averaging the contrasts of the local windows corresponding to all pixels of the input image to obtain the local contrast of the input image.

4. The microscopic image optimization method based on biological visual characteristics according to claim 1, characterized in that: The S1 and S2 retinal models include: The photoreceptor layer is used to simulate the cones and rods in the biological retina and perform local brightness adaptation on the input image; The horizontal cell layer is used to simulate the horizontal cells in the biological retina, providing lateral inhibition signals and performing local contrast enhancement on the output data of the photoreceptor layer; The bipolar cell layer is used to simulate the bipolar cells in the biological retina and further integrate the output data of the horizontal cell layer and enhance the contrast; The ganglion cell layer, used to simulate the ganglion cells in the biological retina, processes the output data of the bipolar cell layer through the central visual channel and the peripheral visual channel.

5. The microscopic image optimization method based on biological visual characteristics according to claim 4, characterized in that: The calculation process expression of the photoreceptor layer is: ， Among them, P adapted (x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t after local brightness adaptive adjustment, ε1 is the first constant used to prevent division by zero, pLAS is the local adaptation sensitivity of the photoreceptor, and P(x, y, t) is the pixel output value at the (x, y) coordinate of the photoreceptor layer at time t without local brightness adaptive adjustment; ， Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, * is the convolution operation, I denoised (x,y,t) is the pixel value at the (x,y) coordinate of the input image at time t.

6. The microscopic image optimization method based on biological visual characteristics according to claim 4, characterized in that: The operation expression of the horizontal cell layer is: ， Among them, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, P adapted (x, y, t) is the (x, y) coordinate pixel output value of the photoreceptor layer after local brightness adaptive adjustment at time t, hCG is the horizontal cell gain, and H(x, y, t) is the (x, y) coordinate value of the lateral inhibition signal at time t; ， Among them, G σS (x,y) is the (x,y) coordinate value of the spatial Gaussian kernel, G σt (t) is the time value of the Gaussian kernel at time t, and * is the convolution operation.

7. The microscopic image optimization method based on biological visual characteristics according to claim 4, characterized in that: The operational expression of the bipolar cell layer is: ， Among them, IPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, B ON_enhanced (x, y, t) is the pixel contrast enhancement value of the (x, y) coordinate of the ON polar cell at time t, B OFF_enhanced (x, y, t) is the (x, y) coordinate pixel contrast enhancement value of the OFF polarity cell at time t; The ON-type polarity cells are used to respond to increased image brightness; The OFF-type polarity cells are used to respond to image brightness reduction; ， ， Among them, B ON (x, y, t) is the (x, y) coordinate value of ON polar cell at time t, B OFF (x, y, t) is the (x, y) coordinate value of the OFF polarity cell at time t, and hCG is the horizontal cell gain; ， ， Where max(,) is the function for finding the maximum value, and OPL(x,y,t) is the pixel output value of the (x,y) coordinate of the horizontal cell layer at time t.

8. The microscopic image optimization method based on biological visual characteristics according to claim 4, characterized in that: The operation process expression of the central visual channel of the ganglion cell layer is: ， Where CV (x, y, t) is the (x, y) coordinate pixel output value of the central visual channel of the ganglion cell layer at time t, OPL (x, y, t) is the (x, y) coordinate pixel output value of the horizontal cell layer at time t, cVGCS is the sensitivity of the central visual channel of the ganglion cell, and ε2 is a second constant used to prevent division by zero; The computational expression of the peripheral visual channel of the ganglion cell layer is: ， Where PV (x, y, t) is the (x, y) coordinate pixel output value of the peripheral visual channel of the ganglion cell layer at time t, ε3 is the third constant used to prevent division by zero, and pVGCS is the sensitivity of the ganglion cell peripheral visual channel. is the operation of finding the partial derivative with respect to time.

9. The microscopic image optimization method based on biological visual characteristics according to claim 1, characterized in that: The S3 includes the following sub-steps: S31, converting the output data of the central visual channel from the BGR color space to the YCrCb color space; S32. Linearly map the output data of the peripheral visual channel of the ganglion cell layer to a value range of 0 to 255 using the following formula, so that it is aligned with the value range of the Y component of the YCrCb color space: ， Among them, P norm is the image data after linear mapping, P is the image data before linear mapping, max() is the function for finding the maximum value, and min() is the function for finding the minimum value; S33, retaining the Cr component and Cb component of the output data of the central visual channel of the ganglion cell layer after conversion to the YCrCb color space, and using the output data of the peripheral visual channel of the ganglion cell layer after linear mapping as a new Y component, and fusing them to obtain an optimized image in the YCrCb space; S34: Convert the optimized image in the YCrCb space back to the BGR space to obtain the final optimized image.

Images