Noise reduction method and system for esophageal optical coherence tomography image

Abstract

The invention relates to the technical field of medical image processing, in particular to a noise reduction method and system for an esophageal optical coherence tomography image and a computer readable storage medium, and the method comprises the steps: extracting the edge of an esophageal OCT original image, and obtaining an edge image; respectively carrying out wavelet transformation on the original image and the edge image to obtain a wavelet transformation coefficient, converting the wavelet transformation coefficient into a column vector form, and obtaining a wavelet coefficient vector and an edge indication vector; constructing an optimization equation of the wavelet coefficient vector by using a sparse optimization algorithm, and performing iterative solution by using the edge indication vector to obtain an optimized wavelet coefficient vector; and obtaining an esophagus OCT noise reduction image according to the optimized wavelet coefficient vector. According to the method, the tissue layer edge information of the esophageal OCT image is reserved while image noise reduction is achieved, more accurate data and edge features are provided for esophageal image analysis and related esophageal disease diagnosis, and the accuracy of esophageal disease diagnosis is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for reducing noise of esophageal optical coherence tomography images and a computer-readable storage medium. Background Art

[0002] With the change of people's lifestyle and eating habits, the incidence of esophageal diseases has increased year by year, seriously endangering people's life and health. At present, esophageal diseases are clinically detected by endoscopy, but this method is expensive and inconvenient to operate, and may also cause serious misdiagnosis due to inaccurate sampling.

[0003] Existing research combines optical coherence tomography (OCT) technology with endoscopy to achieve real-time imaging of internal organs and tissues of the human body, providing a new means for the detection of esophageal diseases such as Barrett's esophagus and eosinophilic esophagitis.

[0004] Although the high-resolution optical images provided by OCT endoscopy technology provide a good means for the early diagnosis and treatment of gastrointestinal tumors, OCT imaging will introduce speckle noise into the image, causing serious damage to the original continuous and clear tissue structure, thus causing great difficulties for subsequent manual or automatic identification, diagnosis, processing and other operations. Therefore, it is necessary to design an effective noise reduction scheme to suppress the noise generated by OCT imaging, so as to help quantify and visualize related structures and pathological information, and achieve accurate diagnosis of esophageal diseases. Effective esophageal endoscopy OCT image noise reduction technology will significantly improve the information of early gastrointestinal mucosal lesions, making accurate diagnosis of early gastrointestinal cancer possible.

[0005] There are many types of noise in OCT images, among which the most serious and largest impact on image quality is multiplicative speckle noise caused by the optical coherence used in OCT imaging technology itself. Compared with traditional additive noise, multiplicative noise is more difficult to remove. In addition, the effective segmentation of esophageal tissue layers is the focus of subsequent processing tasks of esophageal OCT images, which requires that the designed noise reduction method must consider maintaining or enhancing the sharpness of tissue edges in the image, and then distinguishing adjacent tissues.

[0006] In the image denoising task, multiple technologies can be used, such as median filtering, Gaussian filtering, wavelet denoising, diffusion filtering, etc. Among them, median filtering and Gaussian filtering are convenient to implement and have high processing efficiency, but it is difficult to maintain edge information during the denoising process of esophageal OCT images, which is not conducive to subsequent analysis.

[0007] Diffusion filtering is a commonly used model in medical image processing, and the most representative one is the anisotropic diffusion model (PM model). On this basis, researchers have made improvements and proposed a complex diffusion filter model. Other researchers have compared the application of the PM model and the complex diffusion filter in retinal OCT image denoising and enhancement, and found that the complex diffusion filter has better edge enhancement capabilities. However, the complex diffusion filter uses iterative calculations to gradually denoise, so its main disadvantage is that the calculation is large and the calculation efficiency is low.

[0008] In summary, traditional median filtering and Gaussian filtering usually lead to blurred image edges, thus affecting the subsequent analysis of esophageal OCT images and the diagnosis of related diseases. Algorithms based on complex diffusion filtering have disadvantages such as large amount of calculation and unstable edge preservation performance. Therefore, how to improve the denoising effect and efficiency of esophageal OCT images is an urgent problem to be solved. Summary of the invention

[0009] Therefore, the technical problem to be solved by the present invention is to overcome the problems of low denoising effect and efficiency of esophageal OCT images in the prior art.

[0010] In order to solve the above technical problems, the present invention provides a method for reducing noise of esophageal optical coherence tomography images, comprising:

[0011] Obtaining the original OCT image of the esophagus, extracting the edge of the original image, and obtaining an edge image;

[0012] Perform wavelet transform on the original image and edge image respectively, convert the wavelet transform coefficient of the original image into column vector form, recorded as wavelet coefficient vector; convert the wavelet coefficient of the edge image into column vector form, recorded as edge indicator vector;

[0013] Using a sparse optimization algorithm to construct an optimization equation for a wavelet coefficient vector, and using an edge indicator vector to iteratively solve the optimization equation to obtain an optimized wavelet coefficient vector;

[0014] According to the optimized wavelet coefficient vector, the esophageal OCT denoised image is obtained by inverse wavelet transform.

[0015] Preferably, the Sobel operator is used to extract the edge of the original image to obtain an edge image.

[0016] Preferably, the Sobel operator is used to extract the edge of the original image, including:

[0017] The original image is convolved using the horizontal and vertical convolution kernels of the Sobel operator to calculate the horizontal and vertical gradients of each pixel in the original image respectively;

[0018] According to the horizontal gradient and vertical gradient of each pixel in the original image, the gradient magnitude and direction of each pixel are calculated. The formula is:

[0019]

[0020] θ=arctan(G x / G y )

[0021] Among them, I represents the original image, G x represents the horizontal gradient, G y represents the vertical gradient, represents the gradient magnitude, and θ represents the gradient direction.

[0022] Preferably, a dual-tree complex wavelet function is used to perform wavelet transform on the original image and the edge image respectively.

[0023] Preferably, a sparse optimization algorithm is used to construct an optimization equation for the wavelet coefficient vector, including:

[0024]

[0025] in, represents the optimized wavelet coefficient vector, α represents the wavelet coefficient vector, I represents the original image, λ is the sparsity parameter, and A represents the inverse wavelet transform.

[0026] Preferably, an alternating direction multiplier algorithm is used to iteratively solve the optimization equation to obtain an optimized wavelet coefficient vector.

[0027] Preferably, the formula for iteratively solving the optimization equation is:

[0028]

[0029] Among them, A H represents the wavelet forward transform, μ is a constant, α represents the wavelet coefficient vector, I represents the original image, y and z represent arbitrarily set initial vectors, t represents the number of iterations, shrink represents the soft threshold, and α i represents the i-th value of the wavelet coefficient vector α, y i and z i denotes the i-th value of the initial vectors y and z, respectively, and E i represents the i-th value of the edge indicator vector E;

[0030] When the maximum number of iterations is reached or |α t+1 -α t |1 is less than the preset tolerance, the iteration stops, and the α at this time is output t+1 As the optimized wavelet coefficient vector

[0031] Preferably, the calculation formula of the soft threshold is:

[0032]

[0033] in,

[0034] The present invention also provides a noise reduction system for esophageal optical coherence tomography images, comprising:

[0035] An edge extraction module is used to obtain the original esophageal OCT image, extract the edge of the original image, and obtain an edge image;

[0036] The wavelet transform module is used to perform wavelet transform on the original image and the edge image respectively to obtain wavelet transform coefficients; the wavelet transform coefficients of the original image are converted into a column vector form, which is recorded as a wavelet coefficient vector; the wavelet coefficients of the edge image are converted into a column vector form, which is recorded as an edge indicator vector;

[0037] A sparse optimization module, used for constructing an optimization equation of a wavelet coefficient vector using a sparse optimization algorithm, and iteratively solving the optimization equation using an edge indicator vector to obtain an optimized wavelet coefficient vector;

[0038] The noise reduction image acquisition module is used to obtain the noise reduction image of the esophagus OCT through inverse wavelet transform according to the optimized wavelet coefficient vector.

[0039] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for reducing noise of esophageal optical coherence tomography images are implemented.

[0040] The above technical solution of the present invention has the following beneficial effects compared with the prior art:

[0041] The denoising method of an esophageal optical coherence tomography image described in the present invention extracts an edge image from the esophageal OCT original image, marks the edge position of the image, and further converts the wavelet transform coefficients of the original image and the edge image into a column vector form to achieve marking of edge information in the wavelet coefficient dimension. Finally, in a sparse optimization process, the wavelet coefficients of the original image are optimized using the edge information to retain the pixel intensity at the edge position, that is, optimization is performed only at non-edge locations, thereby retaining the tissue layer edge information of the esophageal OCT image while achieving image denoising, providing more accurate data and edge features for esophageal image analysis and related esophageal disease diagnosis, and improving the accuracy of esophageal disease diagnosis.

[0042] In addition, based on the characteristic that the tissue layers of the esophageal image are horizontally structured, the present invention improves the Sobel operator when extracting the edge image using the esophageal OCT original image, and uses the vertical gradient that can better reflect the edge of the tissue layer to represent the gradient amplitude of the pixel within the preset gradient angle range, so as to improve the accuracy of edge extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0044] Figure 1 It is a flow chart of a method for reducing noise of esophageal optical coherence tomography images of the present invention;

[0045] Figure 2 is a schematic diagram of the high-frequency subband direction of the dual-tree complex wavelet, where Figure 2 (a) is a schematic diagram of the 165° high frequency subband. Figure 2 (b) is a schematic diagram of the 135° high frequency subband. Figure 2 (c) is a schematic diagram of the 105° high frequency subband. Figure 2 (d) is a schematic diagram of the 75° high frequency subband. Figure 2 (e) is a schematic diagram of the 45° high frequency subband. Figure 2 (f) is a schematic diagram of the 15° high frequency subband;

[0046] Figure 3 is a set of noise reduction result images obtained in Example 2, where Figure 3 (a) is the original image with speckle noise. Figure 3 (b) in the figure is the denoised image after median filtering. Figure 3 (c) in the figure is the denoised image after wavelet thresholding. Figure 3 (d) in the figure is the noise reduction image of the present invention. DETAILED DESCRIPTION

[0047] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0048] Embodiment 1

[0049] Wavelet denoising methods have achieved good results in both natural image processing and medical image processing. In OCT image processing, Forouzanfer et al. used the directional characteristics of retinal OCT images to study wavelet denoising of retinal OCT images. This method uses the directional characteristics of the edges of retinal OCT images in a macroscopic way, assumes that the layer edges are horizontal, and uses the inherent direction of wavelet decomposition to roughly distinguish edge coefficients from noise coefficients and perform statistics on them. Chitchian et al. achieved good results in denoising prostate OCT images based on dual-tree wavelet transform.

[0050] In recent years, the proposal of sparse expression has attracted widespread attention from researchers. This technology can describe data using as few elements as possible, and can detect the internal structure of data well while processing efficiently. The idea of ​​sparse expression has achieved remarkable research results in signal processing, face recognition, visual tracking, etc. In terms of image denoising, sparse expression has achieved even better results. Elad et al. first combined sparse expression with image denoising and proposed the KSVD denoising method, which is still one of the benchmarks for denoising research. In the field of OCT image denoising, some researchers have also begun to pay attention to sparse expression. For example, Jian et al. designed a dictionary using Curvelet transform to perform sparse expression denoising on OCT images, which significantly improved the image quality. Fang et al. used automatic segmentation information to construct a structural dictionary containing tissue layer information, and then used sparse expression to achieve denoising of retinal OCT images.

[0051] The present invention further considers the characteristic that the edges of esophageal OCT images are generally in the horizontal direction, and proposes a noise reduction method for esophageal optical coherence tomography images. Figure 1 As shown, including:

[0052] S1. Obtain an original OCT image of the esophagus, extract the edge of the original image, and obtain an edge image.

[0053] This embodiment uses the Sobel operator to extract the image boundary of the original esophageal OCT image to assist in subsequent noise reduction.

[0054] The Sobel operator is a classic edge detection algorithm, which is often used in image processing. Its core idea is to use the convolution kernel to calculate the grayscale change of each pixel in the image in the horizontal and vertical directions, that is, the gradient, so as to find the edge in the image.

[0055] In this embodiment, the Sobel operator includes two 3×3 Sobel convolution kernels, and the Sobel convolution kernels in the horizontal direction and the vertical direction are respectively:

[0056]

[0057] Among them, S x is the Sobel convolution kernel in the horizontal direction, S y is the Sobel convolution kernel in the vertical direction.

[0058] The Sobel operator is used to convolve the original image and the horizontal gradient and vertical gradient of each pixel in the original image are calculated respectively. The formula is:

[0059] G x =S x *I,G y =S y *I

[0060] Among them, I represents the original image, G x represents the horizontal gradient, G y Represents the vertical gradient.

[0061] According to the horizontal gradient and vertical gradient of each pixel in the original image, the gradient magnitude and direction of each pixel are calculated. The formula is:

[0062]

[0063] in, represents the gradient magnitude, which reflects the edge strength of each pixel; θ represents the gradient direction.

[0064] Since the tissue layers of the esophageal image usually present a horizontal structure, and the edges of the esophageal OCT image are usually horizontal, the change of the vertical gradient in the image can better reflect the edge of the tissue layer. Therefore, for the part where the gradient direction |θ|>30°, this embodiment will remove G x Component, the expression of the target gradient amplitude is:

[0065]

[0066] Finally, pixels with target gradient amplitude greater than a preset threshold are regarded as the edge of the esophageal tissue to obtain an edge image.

[0067] S2. Perform wavelet transform on the original image and edge image respectively to obtain wavelet transform coefficients; convert the wavelet transform coefficients of the original image into column vector form, recorded as wavelet coefficient vector; convert the wavelet coefficients of the edge image into column vector form, recorded as edge indicator vector.

[0068] The sparse expression of esophageal OCT images requires that the dictionary can describe the edge structure characteristics within the image, so that the noise reduction process can distinguish between the edge and the internal area of ​​the tissue layer, and enhance the useful information while reducing noise. Wavelet transform can accurately reflect the local characteristics of the signal through the expansion and translation of the wavelet function, and is a commonly used transformation dictionary.

[0069] The traditional two-dimensional discrete wavelet transform can generate one low-frequency sub-band and three high-frequency sub-bands by decomposing the image once. The three high-frequency sub-bands represent the detailed structure of the image, corresponding to the horizontal, vertical and diagonal directions in the image respectively.

[0070] However, the tissue structure in the esophageal OCT image is relatively complex, and only three directions cannot well describe all the esophageal tissue layer information. To solve this problem, the present invention intends to use a dual-tree complex wavelet function with more directional selectivity to establish a dictionary to encode the original image and the edge image.

[0071] The dual-tree complex wavelet can be expressed as:

[0072] φ(t)=φ τ (t)+φ l (t)

[0073] Among them, t represents the one-dimensional time domain, φ τ (t) and φ l (t) represent the real and imaginary parts of the dual-tree complex wavelet, both of which are real functions.

[0074] The two-dimensional dual-tree complex wavelet transform can be obtained by the tensor product of the one-dimensional dual-tree complex wavelet, expressed as:

[0075] φ(x,y)=φ(x)φ(y)

[0076] φ(x)=φ τ (x)+jφ l (x), φ(y)=φ τ (y)+jφ l (y)

[0077] Among them, (x, y) represents the two-dimensional spatial domain, φ(x) represents the wavelet transform of the image in the horizontal direction, and φ(y) represents the wavelet transform of the image in the vertical direction. Based on the two-dimensional dual-tree complex wavelet transform φ(x, y), the image can be decomposed into a low-frequency subband matrix and multiple high-frequency subband matrices to obtain the wavelet transform coefficients.

[0078] All subband matrices of the original image are expanded into one-dimensional vector form and concatenated together to form a wavelet coefficient vector. All subband matrices of the edge image are expanded into one-dimensional vector form and concatenated together to form an edge indicator vector.

[0079] The direction corresponding to the high frequency subband of the dual-tree complex wavelet is as follows Figure 2 As shown, Figure 2 (a) is a schematic diagram of the 165° high frequency subband. Figure 2 (b) is a schematic diagram of the 135° high frequency subband. Figure 2(c) is a schematic diagram of the 105° high frequency subband. Figure 2 (d) is a schematic diagram of the 75° high frequency subband. Figure 2 (e) is a schematic diagram of the 45° high frequency subband. Figure 2 (f) in the figure is a schematic diagram of the 15° high-frequency subband. The multi-directional wavelet can form a good match with the edge of the esophageal tissue layer. Using this wavelet to construct a dictionary for expression can effectively reflect the edge information inside the esophageal OCT image, creating favorable conditions for esophageal OCT noise reduction.

[0080] The wavelet transform coefficients of the original image and the edge image are respectively expanded and arranged into one-dimensional vector forms to obtain a wavelet coefficient vector and an edge indicator vector, thereby marking the edge information in the wavelet coefficient dimension.

[0081] S3. Use a sparse optimization algorithm to construct an optimization equation for a wavelet coefficient vector, and use an edge indicator vector to iteratively solve the optimization equation to obtain an optimized wavelet coefficient vector.

[0082] In order to achieve denoising using sparse optimization, the optimization equation of the wavelet coefficient vector is constructed using the sparse optimization algorithm:

[0083]

[0084] in, represents the optimized wavelet coefficient vector, α represents the wavelet coefficient vector, I represents the original image, λ is the sparsity parameter, and A represents the inverse wavelet transform. ||α||1=∑|α i |, used to measure the sparsity of the wavelet coefficient vector α, where α i is the i-th value in the wavelet coefficient vector.

[0085] In order to comprehensively consider the accuracy and efficiency of denoising, the present invention adopts the alternating direction method of multipliers (ADMM) algorithm to iteratively solve the optimization equation, and introduces the edge indicator vector E in the solution process, so that the denoising process is only implemented on the non-edge area and the wavelet coefficients of the edge area are retained.

[0086] The iterative solution process is as follows:

[0087]

[0088] Among them, A H represents the wavelet forward transform, μ is a constant, y and z represent arbitrarily set initial vectors, t represents the number of iterations, and y i and z i denotes the i-th value of the initial vectors y and z, respectively, and E i Represents the i-th value of the edge indicator vector E.

[0089] Shrink represents the soft threshold, and the calculation formula is:

[0090]

[0091] in,

[0092] When the maximum number of iterations is reached or |α t+1 -α t |1 is less than the preset allowable error, the iteration stops, and the optimized wavelet coefficient vector is output Optimized wavelet coefficient vector It can suppress the influence of noise without losing edge information.

[0093] S4. According to the optimized wavelet coefficient vector, the esophageal OCT noise reduction image is obtained by inverse wavelet transform. The formula is:

[0094]

[0095] in, This is a noise-reduced OCT image of the esophagus.

[0096] Embodiment 2

[0097] In order to verify the effectiveness of the esophageal optical coherence tomography image noise reduction method described in the present invention, this embodiment performs a noise reduction experiment on esophageal OCT images.

[0098] The esophageal OCT data used in this example were collected from 5 healthy male black mice, and a total of 50 OCT B-scan data were obtained, with a data size of 1024×1024.

[0099] Median filtering and wavelet threshold denoising have good performance in terms of efficiency and denoising effect, and are therefore often used in OCT image denoising. The method of the present invention is compared with median filtering and wavelet threshold denoising, and the parameters of the various denoising methods used are as follows:

[0100] (1) Median filtering: The filter template size is 5×5;

[0101] (2) Wavelet threshold denoising: The third-order Daubechies wavelet is used as the wavelet, and a total of four layers of wavelet transform are performed, and the threshold size is set to 0.5;

[0102] (3) The present invention: selects dual-tree complex wavelet, performs 4-layer wavelet transform, and sets the sparsity parameter λ to 2.

[0103] In order to numerically evaluate the processing results of different methods, this embodiment adopts the following evaluation indicators:

[0104] (1) Mean-square error (MSE): used to describe the similarity between an image and the original image. The smaller the value, the better. The definition is as follows:

[0105]

[0106] Where I is the original image, It is a denoised image with an image size of M×N.

[0107] (2) Signal-to-noise ratio (SNR): It is used to describe the noise level of an image. The larger the value, the better the image quality. It is defined as follows:

[0108]

[0109] (3) Peak signal-to-noise ratio (PSNR): Another indicator that describes the intensity of image noise. It is defined as:

[0110]

[0111] (4) β index: It is used to describe the edge preservation state of the image after denoising. The smaller the index is, the better the edge preservation performance of the algorithm is. It is defined as follows:

[0112]

[0113] Here, Δ represents a high-pass filter.

[0114] A set of noise reduction results obtained in this embodiment are as follows Figure 3 As shown, Figure 3 (a) is the original image with speckle noise. Figure 3 (b) in the figure is the denoised image after median filtering. Figure 3 (c) in the figure is the denoised image after wavelet thresholding. Figure 3 (d) in the figure is the denoised image of the present invention. It can be seen that the original image contains a lot of speckle noise. The median filter can suppress the speckle noise, but the edge of the tissue layer is also blurred, which is not conducive to the subsequent segmentation of the image. The wavelet threshold denoising has less noise than the median filter result, and the edge is relatively clear. The present invention has the best visual effect among the three denoised images, that is, the image has the least noise and the clearest edge.

[0115] The above results can be more accurately presented through numerical indicators. The three methods were used to perform noise reduction on the 50 OCT B-scans obtained, and the statistical results shown in Table 1 were obtained. Among them, the best results are marked in bold.

[0116] Table 1 Numerical indicators of denoising results of different methods

[0117] method MSE SNR PSNR β Median Filter 198.27±13.21 69.39±2.48 25.16±1.23 0.032±0.003 Wavelet Threshold Denoising 138.56±10.17 70.71±2.11 26.71±1.08 0.263±0.005 The present invention 79.17±6.48 73.12±1.97 29.12±1.31 0.292±0.005

[0118] It can be clearly seen from Table 1 that the present invention has the lowest MSE value, which means that the denoised image obtained is closest to the original image. In addition, the results of the present invention have the highest SNR and PSNR, which means that the denoised image obtained contains the least noise. Finally, the present invention has the highest β index, which shows that the edge preservation characteristics of the denoising result of the present invention are the best. In summary, the denoising method of the present invention can effectively improve the existing methods, improve the final denoising effect, and create favorable conditions for subsequent segmentation.

[0119] In summary, the denoising method of an esophageal optical coherence tomography image described in the present invention extracts an edge image from the esophageal OCT original image, marks the edge position of the image, and further converts the wavelet transform coefficients of the original image and the edge image into a column vector form to achieve marking of edge information in the wavelet coefficient dimension. Finally, in the sparse optimization process, the wavelet coefficients of the original image are optimized using the edge information to retain the pixel intensity at the edge position, that is, optimization is performed only at the non-edge, thereby retaining the tissue layer edge information of the esophageal OCT image while achieving image denoising, providing more accurate data and edge features for esophageal image analysis and related esophageal disease diagnosis, and improving the accuracy of esophageal disease diagnosis.

[0120] In addition, based on the characteristic that the tissue layers of the esophageal image are horizontally structured, the present invention improves the Sobel operator when extracting the edge image using the esophageal OCT original image, and uses the vertical gradient that can better reflect the edge of the tissue layer to represent the gradient amplitude of the pixel within the preset gradient angle range, so as to improve the accuracy of edge extraction.

[0121] Embodiment 3

[0122] Based on the esophageal optical coherence tomography image noise reduction method described in Example 1, this embodiment provides an esophageal optical coherence tomography image noise reduction system, including:

[0123] An edge extraction module is used to obtain the original esophageal OCT image, extract the edge of the original image, and obtain an edge image;

[0124] The wavelet transform module is used to perform wavelet transform on the original image and the edge image respectively to obtain wavelet transform coefficients; the wavelet transform coefficients of the original image are converted into a column vector form, which is recorded as a wavelet coefficient vector; the wavelet coefficients of the edge image are converted into a column vector form, which is recorded as an edge indicator vector;

[0125] A sparse optimization module, used for constructing an optimization equation of a wavelet coefficient vector using a sparse optimization algorithm, and iteratively solving the optimization equation using an edge indicator vector to obtain an optimized wavelet coefficient vector;

[0126] The noise reduction image acquisition module is used to obtain the noise reduction image of the esophagus OCT through inverse wavelet transform according to the optimized wavelet coefficient vector.

[0127] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for reducing noise of esophageal optical coherence tomography images are implemented.

[0128] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0132] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for denoising esophageal optical coherence tomography images, characterized in that: include: Obtaining the original OCT image of the esophagus, extracting the edge of the original image, and obtaining an edge image; Perform wavelet transform on the original image and edge image respectively, convert the wavelet transform coefficient of the original image into column vector form, recorded as wavelet coefficient vector; convert the wavelet coefficient of the edge image into column vector form, recorded as edge indicator vector; Using a sparse optimization algorithm to construct an optimization equation for a wavelet coefficient vector, and using an edge indicator vector to iteratively solve the optimization equation to obtain an optimized wavelet coefficient vector; According to the optimized wavelet coefficient vector, the esophageal OCT denoised image is obtained by inverse wavelet transform.

2. The method for denoising an esophageal optical coherence tomography image according to claim 1, characterized in that: The Sobel operator is used to extract the edge of the original image to obtain an edge image.

3. The method for reducing noise of esophageal optical coherence tomography images according to claim 2, characterized in that: The Sobel operator is used to extract the edge of the original image, including: The original image is convolved using the horizontal and vertical convolution kernels of the Sobel operator to calculate the horizontal and vertical gradients of each pixel in the original image respectively; According to the horizontal gradient and vertical gradient of each pixel in the original image, the gradient magnitude and direction of each pixel are calculated. The formula is: θ=arctane(G x / G y ) Among them, I represents the original image, G x represents the horizontal gradient, G y represents the vertical gradient, represents the gradient magnitude, and θ represents the gradient direction.

4. The method for denoising esophageal optical coherence tomography images according to claim 1, characterized in that: The dual-tree complex wavelet function is used to perform wavelet transform on the original image and the edge image respectively.

5. The method for denoising esophageal optical coherence tomography images according to claim 1, characterized in that: The optimization equation of the wavelet coefficient vector is constructed using a sparse optimization algorithm, including: in, represents the optimized wavelet coefficient vector, α represents the wavelet coefficient vector, I represents the original image, λ is the sparsity parameter, and A represents the inverse wavelet transform.

6. The method for denoising esophageal optical coherence tomography images according to claim 5, characterized in that: The optimization equation is iteratively solved using the alternating direction multiplier algorithm to obtain the optimized wavelet coefficient vector.

7. The method for reducing noise of esophageal optical coherence tomography images according to claim 6, characterized in that: The formula for iterative solution of the optimization equation is: Among them, A H represents the wavelet forward transform, μ is a constant, α represents the wavelet coefficient vector, I represents the original image, y and z represent arbitrarily set initial vectors, t represents the number of iterations, shrink represents the soft threshold, and α i represents the i-th value of the wavelet coefficient vector α, y i and z i denote the i-th value of the initial vectors y and z, respectively, and E i represents the i-th value of the edge indicator vector E; When the maximum number of iterations is reached or |α t+1 -α t |1 is less than the preset tolerance, the iteration stops, and the α at this time is output t+1 As the optimized wavelet coefficient vector 8. The method for reducing noise of esophageal optical coherence tomography images according to claim 7, characterized in that: The calculation formula of the soft threshold is: in, 9. A noise reduction system for esophageal optical coherence tomography images, characterized in that: include: An edge extraction module is used to obtain the original esophageal OCT image, extract the edge of the original image, and obtain an edge image; The wavelet transform module is used to perform wavelet transform on the original image and the edge image respectively to obtain wavelet transform coefficients; the wavelet transform coefficients of the original image are converted into a column vector form, which is recorded as a wavelet coefficient vector; the wavelet coefficients of the edge image are converted into a column vector form, which is recorded as an edge indicator vector; A sparse optimization module, used for constructing an optimization equation of a wavelet coefficient vector using a sparse optimization algorithm, and iteratively solving the optimization equation using an edge indicator vector to obtain an optimized wavelet coefficient vector; The noise reduction image acquisition module is used to obtain the noise reduction image of the esophagus OCT through inverse wavelet transform according to the optimized wavelet coefficient vector.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for reducing noise of esophageal optical coherence tomography images as claimed in any one of claims 1 to 8 are implemented.

Images