An adaptive weighted bilateral filter deconvolution method and system for OCT image processing

Abstract

The application relates to an OCT image processing method and system based on adaptive weighted bilateral filtering deconvolution, which comprises the following steps: preprocessing an original OCT signal, including removing background noise, wave number calibration, spectrum shaping and dispersion compensation, to obtain an OCT reconstructed image; adjusting the spatial weight and intensity weight of the OCT stretched image through histogram stretching to enhance contrast, and adaptively denoising the image; further optimizing the image by using a blind deconvolution algorithm, iteratively updating the estimated image and a point spread function (PSF), and obtaining a high-definition OCT reconstructed image. By using the adaptive weighted bilateral filtering and blind deconvolution technology, the application realizes efficient denoising and detail reservation of the OCT image, thereby obtaining a high-definition reconstructed image.

Description

Technical Field

[0001] This invention relates to the field of OCT image processing, and specifically to an adaptive weighted bilateral filtering deconvolution OCT image processing method. Background Technology

[0002] The point spread function (PSF) describes the distribution of light spots formed on the imaging plane after an ideal point light source passes through an optical system. It is used to evaluate and improve the quality of imaging systems and is a key tool in image restoration, deblurring, and optical system design and calibration.

[0003] Blind deconvolution is an image processing technique that recovers a sharp original image from a blurred image without knowing or fully knowing the point spread function (PSF). Unlike traditional deconvolution methods that require precise knowledge of the PSF, blind deconvolution algorithms can simultaneously estimate the PSF and recover image content. This makes it particularly suitable for practical applications where it is difficult to directly measure the PSF, such as astronomical observation, microscopic imaging, and general image restoration tasks. This method can effectively improve image quality and detail sharpness.

[0004] In the field of medical imaging, optical coherence tomography (OCT) is an important non-invasive imaging technique that can provide high-resolution three-dimensional images of tissue structures. However, due to the coherence of light during the imaging process, OCT images are often affected by speckle noise. Speckle noise is a type of random noise caused by coherent scattering during the imaging process, which significantly reduces image contrast and sharpness, affecting doctors' accurate identification and diagnosis of lesions. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes an adaptive weighted bilateral filtering deconvolution method for OCT images. By combining adaptive bilateral filtering with blind deconvolution techniques, the spatial and intensity kernel parameters of the filter are dynamically adjusted, effectively improving the processing quality of OCT images, enhancing the structural and detail integrity of the images, and providing technical support for clinical OCT imaging.

[0006] The specific plan is as follows:

[0007] On the one hand, an adaptive weighted bilateral filtering deconvolution OCT image processing method is characterized by comprising the following steps:

[0008] S1, preprocess the original tissue OCT signal to obtain the OCT reconstructed image;

[0009] S2, Histogram stretching is performed on the OCT reconstructed image to obtain the OCT stretched image;

[0010] S3, adjust the spatial weights and intensity weights of the OCT stretched image; based on the adjusted spatial weights and intensity weights, perform adaptive denoising on the OCT stretched image to obtain the OCT denoised image;

[0011] S4 takes the OCT denoised image and the initialized point spread function PSF as input, performs a blind deconvolution algorithm, and obtains the final OCT output image.

[0012] Furthermore, the preprocessing specifically includes:

[0013] Background noise removal, wavenumber calibration, spectral shaping, and dispersion compensation were performed on the original tissue OCT signal.

[0014] Furthermore, the histogram stretching specifically includes:

[0015] Determine the lowest non-zero intensity value L from the OCT reconstructed image. ′ min and the highest non-zero intensity value L ′ max The linear intensity transformation is applied to map the original intensity value I(x,y) of each pixel in the OCT reconstructed image to a new pixel value p(i,j); the formula for the linear intensity transformation is as follows:

[0016]

[0017] Where x is the horizontal coordinate of the image, representing the position of the pixel in the horizontal direction; y is the vertical coordinate of the image, representing the position of the pixel in the vertical direction; i is the horizontal coordinate of the transformed image, and j is the vertical coordinate of the transformed image; L max and L min These represent the maximum and minimum intensity values ​​in the original OCT reconstructed image, respectively;

[0018] The pixel value of each pixel in the image after linear intensity transformation is counted, and the number of pixels at each intensity level is accumulated to complete the histogram stretching.

[0019] Furthermore, the intensity weight is the actual weight calculated by the intensity kernel, which measures the intensity difference between two pixels in an image. The intensity weight is the output value of the intensity kernel; the formula for calculating the intensity weight is as follows:

[0020] σ r (m,n)=βμ p

[0021]

[0022] Where β represents the coefficient for adjusting the magnitude of the standard deviation of strength; μp σ represents the local mean; m and n represent the x and y coordinates of the current center pixel, which are the core reference points for weight calculation; r (m,n) represents the local strength standard deviation based on the center point (m,n); w r (m,n) represents the intensity weight; k and l represent the horizontal and vertical coordinates of a pixel being traversed within the neighborhood window; I(m,n) represents the intensity value of the currently being processed center pixel; I(k,l) represents the intensity value of a point in the neighborhood.

[0023] Furthermore, the spatial weights are actual weights calculated from a spatial kernel, which is a function used to measure the similarity between two pixels in the image in terms of spatial location. The spatial weights represent the output value of the spatial kernel; first, the local entropy value H[I] is used. p Calculate σ s After calculating the spatial standard deviation, the spatial weights are then calculated using the following formula:

[0024]

[0025] Where p(i) represents the probability of pixel value i appearing in the neighborhood; B N This represents the total number of pixel values ​​within the neighborhood; and These represent the maximum and minimum preset spatial standard deviations, respectively; ζ represents a preset threshold used to adjust the range of variation in the spatial kernel size; H[I p ] represents the entropy of a local region of the image; I represents the intensity value of the image; I p This represents the pixel intensity value of a local region in an image; w s (u,v) represents the spatial weights, where u and v represent the x and y coordinates of the center pixel, respectively; s and t represent the x and y coordinates of a pixel traversed within the neighborhood window; σ s (u,v) represents the local spatial standard deviation based on the center point (u,v).

[0026] Furthermore, the adaptive denoising process is based on the calculated intensity weight w. r and spatial weight w s Bilateral filtering is applied to the OCT reconstructed image after histogram stretching. The filtering operation slides across the image pixel by pixel until the entire image is processed, resulting in an OCT reconstructed image that preserves image edge and detail information. The specific formula is as follows:

[0027] w h (i,j)=w r (i,j)·w s (i,j)

[0028]

[0029] Where i represents the x-coordinate of the image after histogram stretching, and j represents the y-coordinate of the image after transformation; b This represents the final filtered output image; This indicates that a bilateral filtering operation is performed using specific spatial and intensity standard deviations. and represent the minimum and maximum values ​​of the spatial standard deviation, respectively, used for optimization targeting different image features; weight w h The final weight is calculated based on the intensity weight and the spatial weight.

[0030] Furthermore, the blind deconvolution algorithm is executed to estimate the OCT denoised image and optimize the point spread function (PSF), and the calculation formula is as follows:

[0031]

[0032] Among them, I out [k] represents the image estimate after the k-th iteration; I obs This represents the observed denoised OCT reconstructed image; H represents the estimated point spread function (PSF). λ represents the inverse or pseudo-inverse of the point spread function (PSF); λ represents the iteration step size, used to control the update speed.

[0033] On the other hand, an adaptive weighted bilateral filtering deconvolution OCT image processing system is characterized by comprising:

[0034] The preprocessing module is used to preprocess the original tissue OCT signal to obtain an OCT reconstructed image;

[0035] The histogram stretching module is used to stretch the histogram of the OCT reconstructed image to obtain a stretched OCT image.

[0036] An adaptive denoising module is used to adjust the spatial and intensity weights of the OCT stretched image; based on the adjusted spatial and intensity weights, adaptive denoising processing is performed on the OCT stretched image to obtain a denoised OCT image.

[0037] The blind deconvolution module takes the OCT denoised image and the initialized point spread function (PSF) as input, executes the blind deconvolution algorithm, and obtains the final OCT output image.

[0038] The present invention adopts the above technical solution and has the following beneficial effects:

[0039] (1) Through adaptive weighted bilateral filtering technology, this invention can effectively remove speckle noise and other Gaussian white noise in OCT images, and significantly improve the contrast and clarity of the images;

[0040] (2) This invention utilizes histogram stretching and local statistical properties to adjust spatial weights and intensity weights, preserving important edge and detail information during the denoising process, thereby enhancing the visual effect and diagnostic value of the image.

[0041] (3) This invention combines blind deconvolution algorithm, optimizes point spread function PSF, and estimates the original unblurred image through iterative update, and finally obtains high-quality, high-resolution OCT reconstructed image, which helps to more accurate medical analysis. Attached Figure Description

[0042] Figure 1 This is a flowchart of the OCT image processing method using adaptive weighted bilateral filtering deconvolution according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram illustrating the implementation of the adaptive weighted bilateral filtering deconvolution OCT image processing method according to an embodiment of the present invention;

[0044] Figure 3 These are the skin OCT noise image and the processed image according to an embodiment of the present invention;

[0045] Figure 4 These are OCT noise images and processed images of the adhesive tape according to an embodiment of the present invention;

[0046] Figure 5 This is a system diagram of the OCT image processing method with adaptive weighted bilateral filtering deconvolution according to an embodiment of the present invention. Detailed Implementation

[0047] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0048] On the one hand, such as Figure 1 and Figure 2 As shown, the present invention provides an adaptive weighted bilateral filtering deconvolution OCT image processing method, which includes the following steps.

[0049] S1, preprocess the original tissue OCT signal to obtain the OCT reconstructed image.

[0050] Specifically, the preprocessing of the raw tissue OCT signal includes:

[0051] Background noise removal, wavenumber calibration, spectral shaping, and dispersion compensation were performed on the original tissue OCT signal.

[0052] Specifically, OCT images are often affected by various types of noise, the most significant of which is speckle noise, generated by multipath reflection and scattering effects. In addition, OCT images may also be affected by laser intensity noise, photoelectron firing noise, and thermal noise generated by electronic devices. These noises are generally considered Gaussian white noise and have a smaller impact than speckle noise, so they can be ignored. Therefore, in this embodiment, OCT image I is defined as... * as follows:

[0053]

[0054] Where I represents a noise-free image, N s N represents speckle noise. w Let Gaussian white noise be represented. Given that Gaussian white noise has a smaller impact than speckle noise, a simplified model is proposed to model the OCT image as follows:

[0055] I * =N s ·I

[0056] A method is proposed to convert multiplicative noise into additive noise, namely by performing a logarithmic transformation on the OCT image, as follows:

[0057] log(I * ) = log(N s )+log(I)

[0058] After the transformation, the noise model is close to the additive model, making noise processing simpler.

[0059] S2, Histogram stretching is performed on the OCT reconstructed image to obtain the OCT stretched image.

[0060] Specifically, the histogram stretching includes:

[0061] Determine the lowest non-zero intensity value L′ from the OCT reconstructed image. min and the highest non-zero intensity value L′ max The linear intensity transformation is applied to map the original intensity value I(x,y) of each pixel in the OCT reconstructed image to a new pixel value p(i,j); the formula for the linear intensity transformation is as follows:

[0062]

[0063] Where x is the horizontal coordinate of the image, representing the position of the pixel in the horizontal direction; y is the vertical coordinate of the image, representing the position of the pixel in the vertical direction; i is the horizontal coordinate of the transformed image, and j is the vertical coordinate of the transformed image; L max and L min These represent the maximum and minimum intensity values ​​in the original OCT reconstructed image, respectively;

[0064] The OCT-reconstructed image processed by linear intensity transformation has improved contrast; subsequently, the pixel value of each pixel in the image is counted, and the number of pixels at each intensity level is accumulated to form a histogram, where a predetermined intensity range (L... ′ max ,L ′ min This is based on the image display or processing requirements that are set in advance.

[0065] Specifically, before applying the linear intensity transform, the color space of the OCT reconstructed image is converted from the RGB color space to the grayscale space; before outputting the OCT reconstructed image after histogram stretching, the image processed by the linear intensity transform is converted from the grayscale space back to the RGB color space.

[0066] Specifically, local statistical properties refer to local contrast, which measures the degree of difference in brightness within a given area. For OCT images, local contrast can help identify the boundaries or feature regions of different tissues.

[0067] S3. Based on local statistical characteristics, the spatial weights and intensity weights of the OCT reconstructed image after histogram stretching are adjusted; based on the adjusted spatial weights and intensity weights, adaptive denoising processing is performed on the OCT reconstructed image after histogram stretching to obtain the denoised OCT reconstructed image.

[0068] Specifically, the intensity weight is the actual weight calculated by the intensity kernel, which measures the intensity difference between two pixels in an image. The intensity weight is the output value of the intensity kernel. The formula for calculating the intensity weight is as follows:

[0069] σ r (m,n)=βμ p

[0070]

[0071] Where β represents the coefficient for adjusting the magnitude of the standard deviation of strength; μ p σ represents the local mean; m and n represent the x and y coordinates of the current center pixel, which are the core reference points for weight calculation; r (m,n) represents the local strength standard deviation based on the center point (m,n); w r (m,n) represents the intensity weight; k and l represent the horizontal and vertical coordinates of a pixel being traversed within the neighborhood window; I(m,n) represents the intensity value of the currently being processed center pixel; I(k,l) represents the intensity value of a point in the neighborhood.

[0072] Specifically, the spatial weights are the actual weights calculated by the spatial kernel, which is a function used to measure the similarity between two pixels in the image in terms of spatial location. The spatial weights represent the output value of the spatial kernel; first, the local entropy value H[I] is used. p Calculate σ s After calculating the spatial standard deviation, the spatial weights are then calculated using the following formula:

[0073]

[0074] Where p(i) represents the probability of pixel value i appearing in the neighborhood; B N This represents the total number of pixel values ​​within the neighborhood; and These represent the maximum and minimum preset spatial standard deviations, respectively; ζ represents a preset threshold used to adjust the range of variation in the spatial kernel size; H[I p ] represents the entropy of a local region of the image; I represents the intensity value of the image; I p This represents the pixel intensity value of a local region in an image; w s (u,v) represents the spatial weights, where u and v represent the x and y coordinates of the center pixel, respectively; s and t represent the x and y coordinates of a pixel traversed within the neighborhood window; σ s (u,v) represents the local spatial standard deviation based on the center point (u,v).

[0075] Specifically, the adaptive denoising process is based on the calculated intensity weight w r and spatial weight w s Bilateral filtering is applied to the OCT reconstructed image after histogram stretching. The filtering operation slides across the image pixel by pixel until the entire image is processed, resulting in an OCT reconstructed image that preserves image edge and detail information. The specific formula is as follows:

[0076] w h (i,j)=w r (i,j)·w s (i,j)

[0077]

[0078] Where i represents the x-coordinate of the image after histogram stretching, and j represents the y-coordinate of the image after transformation; b This represents the final filtered output image; This indicates that a bilateral filtering operation is performed using specific spatial and intensity standard deviations. and represent the minimum and maximum values ​​of the spatial standard deviation, respectively, used for optimization targeting different image features; weight w hThe final weight is calculated based on the intensity weight and the spatial weight.

[0079] S4 takes the denoised OCT reconstructed image and the initialized point spread function PSF as input, and performs a blind deconvolution algorithm to obtain the final OCT reconstructed image.

[0080] Specifically, the blind deconvolution algorithm is used to estimate the denoised OCT reconstructed image and optimize the point spread function (PSF). The calculation formula is as follows:

[0081]

[0082] Among them, I out [k] represents the image estimate after the k-th iteration; I obs This represents the observed denoised OCT reconstructed image; H represents the estimated point spread function (PSF). λ represents the inverse or pseudo-inverse of the point spread function (PSF); λ represents the iteration step size, used to control the update speed.

[0083] Specifically, the image and PSF estimate are iteratively updated until a predetermined convergence condition is met or a set number of iterations is reached. The deconvolutioned image and optimized PSF are then output, calculated using the following formula:

[0084]

[0085] Finally, the output pixel value I out The image composed of the output pixel values ​​corresponding to pixel point I(x,y) in the original image is the final OCT reconstructed image of the denoising algorithm of this invention.

[0086] Figure 3 and Figure 4 The denoising results for tape and skin are shown below, respectively, using the algorithm. It can be seen that the image denoising method of this invention better reduces noise signals in the image while preserving important details.

[0087] like Figure 5 As shown, this embodiment also discloses an OCT image processing system with adaptive weighted bilateral filtering deconvolution, including:

[0088] Preprocessing module 51 preprocesses the original tissue OCT signal to obtain an OCT reconstructed image;

[0089] Histogram stretching module 52 performs histogram stretching on the OCT reconstructed image to obtain an OCT stretched image;

[0090] The adaptive denoising module 53 is used to adjust the spatial weights and intensity weights of the OCT stretched image; based on the adjusted spatial weights and intensity weights, adaptive denoising processing is performed on the OCT stretched image to obtain an OCT denoised image;

[0091] The blind deconvolution module 54 is used to take the OCT denoised image and the initialized point spread function PSF as input, execute the blind deconvolution algorithm, and obtain the final OCT output image.

[0092] The specific implementation of an adaptive weighted bilateral filtering deconvolution OCT image processing system is the same as that of the adaptive weighted bilateral filtering deconvolution OCT image processing method, and will not be described again in this embodiment.

[0093] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. An OCT image processing method using adaptive weighted bilateral filtering deconvolution, characterized in that, Includes the following steps: S1, preprocess the original tissue OCT signal to obtain the OCT reconstructed image; S2, Histogram stretching is performed on the OCT reconstructed image to obtain the OCT stretched image; S3, adjust the spatial weights and intensity weights of the OCT stretched image; Based on the adjusted spatial and intensity weights, adaptive denoising is performed on the OCT stretched image to obtain the OCT denoised image. S4. Take the OCT denoised image and the initialized point spread function PSF as input, and perform blind deconvolution algorithm to obtain the final OCT output image; The intensity weight is the actual weight calculated by the intensity kernel, which measures the intensity difference between two pixels in an image. The intensity weight is the output value of the intensity kernel. The formula for calculating the intensity weight is as follows: σ r (m, n) = βμ p where β represents a coefficient for adjusting the intensity standard deviation size; μ p represents the local mean; m and n represent the horizontal and vertical coordinates of the current center pixel, which are the core reference points for weight calculation; σ r (m,n) represents the local intensity standard deviation based on the center point (m,n); w r (m,n) represents the intensity weight; k and l represent the horizontal and vertical coordinates of a certain pixel in the neighborhood window; I(m,n) represents the intensity value of the center pixel currently being processed; I(k,l) represents the intensity value of a certain point in the neighborhood. The spatial weight is an actual weight calculated by a spatial kernel, which is a function for measuring the similarity of two pixels in the image in spatial position, and the spatial weight represents the output value of the spatial kernel; the local entropy value H[I p ] is used to calculate σ s spatial standard deviation, and then the spatial weight is calculated, and the formula is as follows: Where p(i) represents the probability of pixel value i appearing in the neighborhood; B N This represents the total number of pixel values ​​within the neighborhood; and These represent the maximum and minimum preset spatial standard deviations, respectively; ζ represents a preset threshold used to adjust the range of variation in the spatial kernel size; H[I p ] represents the entropy of a local region of the image; I represents the intensity value of the image; I p This represents the pixel intensity value of a local region in an image; w s (u,v) represents the spatial weights, where u and v represent the x and y coordinates of the center pixel, respectively; s and t represent the x and y coordinates of a pixel traversed within the neighborhood window; σ s (u,v) represents the local spatial standard deviation based on the center point (u,v); The adaptive denoising process is based on the calculated intensity weight w r and spatial weight w s Bilateral filtering is applied to the OCT reconstructed image after histogram stretching. The filtering operation slides across the image pixel by pixel until the entire image is processed, resulting in an OCT reconstructed image that preserves image edge and detail information. The specific formula is as follows: w h (i,j)=w r (i,j)·w s (i,j) Where i represents the x-coordinate of the image after histogram stretching, and j represents the y-coordinate of the image after transformation; I b This represents the final filtered output image; This indicates that a bilateral filtering operation is performed using specific spatial and intensity standard deviations. and represent the minimum and maximum values ​​of the spatial standard deviation, respectively, used for optimization targeting different image features; weight w h The final weight is calculated based on the intensity weight and the spatial weight.

2. The OCT image processing method with adaptive weighted bilateral filtering and deconvolution according to claim 1, characterized in that, In S1, the preprocessing specifically includes: Background noise removal, wavenumber calibration, spectral shaping, and dispersion compensation were performed on the original tissue OCT signal.

3. The OCT image processing method with adaptive weighted bilateral filtering and deconvolution according to claim 1, characterized in that, In S2, the histogram stretching specifically includes: Determine the lowest non-zero intensity value L from the OCT reconstructed image. ′ min and the highest non-zero intensity value L ′ max The linear intensity transformation is applied to map the original intensity value I(x,y) of each pixel in the OCT reconstructed image to a new pixel value p(i,j); the formula for the linear intensity transformation is as follows: Where x is the horizontal coordinate of the image, representing the position of the pixel in the horizontal direction; y is the vertical coordinate of the image, representing the position of the pixel in the vertical direction; i is the horizontal coordinate of the transformed image, and j is the vertical coordinate of the transformed image; L max and L min These represent the maximum and minimum intensity values ​​in the original OCT reconstructed image, respectively; The pixel value of each pixel in the image after linear intensity transformation is counted, and the number of pixels at each intensity level is accumulated to complete the histogram stretching.

4. The OCT image processing method with adaptive weighted bilateral filtering and deconvolution according to claim 1, characterized in that... In S4, the blind deconvolution algorithm is executed to estimate the OCT denoised image and optimize the point spread function (PSF). The calculation formula is as follows: Among them, I out [k] represents the image estimate after the k-th iteration; I obs This represents the observed denoised OCT reconstructed image; H represents the estimated point spread function (PSF). λ represents the inverse or pseudo-inverse of the point spread function (PSF); λ represents the iteration step size, used to control the update speed.

5. The OCT image processing system based on the adaptive weighted bilateral filtering deconvolution OCT image processing method according to claim 1, characterized in that, include: The preprocessing module is used to preprocess the original tissue OCT signal to obtain the OCT reconstructed image; The histogram stretching module is used to stretch the histogram of the OCT reconstructed image to obtain a stretched OCT image. An adaptive denoising module is used to adjust the spatial and intensity weights of the OCT stretched image; Based on the adjusted spatial weights and intensity weights, adaptive denoising processing is performed on the OCT stretched image to obtain the OCT denoised image. The blind deconvolution module takes the OCT denoised image and the initialized point spread function (PSF) as input, executes the blind deconvolution algorithm, and obtains the final OCT output image.

Images