Search

CN-122023168-A - Image denoising method and device based on Gaussian blur self-adaptive filtering

CN122023168ACN 122023168 ACN122023168 ACN 122023168ACN-122023168-A

Abstract

The invention provides a Gaussian blur-based adaptive filtering image denoising method and device, which comprise the steps of converting an input original image into an integer type to generate a to-be-processed result, performing first Gaussian blur processing on the to-be-processed result to generate a first blur result of a smoothed image, performing second Gaussian blur processing according to the absolute value of the difference value between the to-be-processed result and the first blur result to generate a second blur result, estimating a noise variance according to the second blur result to generate an adaptive filter coefficient, fusing the first blur result and the adaptive filter coefficient, outputting a denoising image of the original image, and updating a noise standard deviation through iteration to adapt to a dynamic noise scene.

Inventors

  • SHI KE
  • YUAN LIANG
  • FANG ZHENGJUN

Assignees

  • 安徽光智科技有限公司

Dates

Publication Date
20260512
Application Date
20251224

Claims (10)

  1. 1. The self-adaptive filtering image denoising method based on Gaussian blur is characterized by comprising the following steps of: Converting an input original image into an integer type and then generating a to-be-processed result; Performing first Gaussian blur processing on the to-be-processed result to generate a primary blur result of the smoothed image; Executing second Gaussian blur processing according to the absolute value of the difference value between the to-be-processed result and the primary blur result to generate a secondary blur result; estimating noise variance according to the secondary blurring result, and generating an adaptive filter coefficient; And fusing the primary blurring result and the self-adaptive filter coefficient, outputting a denoising image of the original image, and adapting to a dynamic noise scene by iteratively updating a noise standard deviation.
  2. 2. The method for denoising an image based on adaptive filtering of gaussian blur according to claim 1, wherein the process of generating a result to be processed after converting an input original image into an integer type is specifically represented as: Wherein, the The original image is represented by a representation of the original image, The coordinates of the pixels are represented and, Representing the result to be processed.
  3. 3. The method for denoising an image based on adaptive filtering of gaussian blur according to claim 2, wherein the process of generating a primary blur result of a smoothed image after performing a first gaussian blur process on the result to be processed specifically comprises: constructing a Gaussian kernel K with a radius r, wherein the Gaussian kernel K is positioned at the position Weight value at Expressed as: The result to be processed Generating a primary blurring result of a smooth image after convolution operation with the gaussian kernel K based on a first gaussian blurring process And is expressed as: in the above, the result is blurred once Represented as pixels The gray value of the smoothed image at that point, Representing the gaussian kernel standard deviation, exp () represents the natural exponential function.
  4. 4. The adaptive filtering image denoising method based on gaussian blur according to claim 3, wherein the process of generating a secondary blur result by performing a convolution operation of a second gaussian blur process according to the absolute value of the difference between the result to be processed and the primary blur result comprises: Generating a residual error value by making a difference between the to-be-processed result and the primary fuzzy result Expressed as: for the residual value Taking absolute value to obtain Expressed as: performing convolution operation of the second Gaussian blur process to generate a secondary blur result Expressed as: In the above, the secondary blurring result Represented as pixels And (5) gray values after secondary Gaussian blur.
  5. 5. The method for denoising an adaptively filtered image based on gaussian blur according to claim 4, wherein estimating noise variance from the secondary blur result specifically comprises: performing noise variance estimation processing according to the secondary blurring result to generate a first noise variance estimation value Expressed as: From the noise variance estimate Standard deviation from initial noise Generating a second noise variance estimation value after noise standard deviation Expressed as: According to the variance estimation value after the noise standard deviation Generating a nonnegatively-charged third noise variance estimate And is expressed as: from the first noise variance estimate Third noise variance estimation value Residual value Generating adaptive filter coefficients Expressed as: In the above-mentioned method, the step of, The normalization factor is represented, and the value is 0, 1.
  6. 6. The method for denoising an image based on adaptive filtering of gaussian blur according to claim 5, wherein the process of fusing the primary blur result and the adaptive filter coefficient and outputting the denoised image of the original image is specifically expressed as: In the above-mentioned method, the step of, Representing the gray value of the denoised image.
  7. 7. The method of claim 5, wherein the step of adapting the dynamic noise scene by iteratively updating the noise standard deviation comprises estimating a first noise variance of the current frame from the de-noised image Generating an updated noise standard deviation for denoising processing of the next frame of image to realize adaptation to a dynamic noise scene; Wherein the updated noise standard deviation Expressed as: In the above-mentioned method, the step of, Representing a first noise variance estimate for a current frame Is used for the histogram statistics of the (c) for the (c), Representing a rounding function for converting the calculation result into an integer.
  8. 8. An image denoising apparatus based on adaptive filtering of gaussian blur to perform the image denoising method according to any one of claims 1 to 7, characterized in that the image denoising apparatus comprises: The to-be-processed module is used for generating a to-be-processed result after converting an input original image into an integer type; The first generation module is used for generating a primary fuzzy result of the smoothed image after executing first Gaussian blur processing on the to-be-processed result; the second generation module is used for executing second Gaussian blur processing according to the absolute value of the difference value between the to-be-processed result and the primary blur result to generate a secondary blur result; the third generation module is used for estimating noise variance according to the secondary blurring result and generating an adaptive filter coefficient; and the denoising output module is used for fusing the primary fuzzy result and the self-adaptive filter coefficient, outputting a denoising image of the original image, and adapting to a dynamic noise scene by iteratively updating the noise standard deviation.
  9. 9. An electronic device comprising a memory and a processor, the memory having stored thereon a computer program, characterized in that the processor, when executing the computer program, implements the method according to any of claims 1-7.
  10. 10. A computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the method according to any of claims 1-7.

Description

Image denoising method and device based on Gaussian blur self-adaptive filtering Technical Field The invention relates to the technical field of image processing, in particular to an image denoising method and device based on self-adaptive filtering of Gaussian blur. Background The image is easily interfered by Gaussian noise, salt and pepper noise and the like in the process of acquisition, transmission or storage, so that the image quality is reduced, and the subsequent analysis and recognition are affected. The existing denoising method mainly comprises the modes of Gaussian filtering, median filtering, self-adaptive filtering, non-local mean filtering and the like, but has the following technical problems: (1) Detail loss, namely, the edge and detail are insufficiently protected by the traditional Gaussian filtering, and the image is blurred due to excessive smoothing; (2) The adaptability is poor, the filtering method with fixed parameters cannot cope with scenes with different noise intensities or types, and the denoising effect is unstable; (3) The noise estimation is inaccurate, the existing algorithm depends on a preset noise model, and when the noise characteristics in the actual scene change dynamically, the denoising precision is reduced; (4) The calculation is complex, a part of high-performance algorithms (such as non-local mean value) need a large amount of redundant calculation, and real-time processing requirements (such as monitoring video streams) are difficult to meet. Disclosure of Invention In order to solve the technical problems, the invention provides an image denoising method and device based on adaptive filtering of Gaussian blur, which can efficiently denoise and keep details by dynamically estimating noise and adaptively adjusting filter coefficients, and has both adaptability and instantaneity. In a first aspect, the present invention provides a method for denoising an image based on adaptive filtering of gaussian blur, which specifically includes the following steps: Converting an input original image into an integer type and then generating a to-be-processed result; Performing first Gaussian blur processing on the to-be-processed result to generate a primary blur result of the smoothed image; Executing second Gaussian blur processing according to the absolute value of the difference value between the to-be-processed result and the primary blur result to generate a secondary blur result; estimating noise variance according to the secondary blurring result, and generating an adaptive filter coefficient; And fusing the primary blurring result and the self-adaptive filter coefficient, outputting a denoising image of the original image, and adapting to a dynamic noise scene by iteratively updating a noise standard deviation. Further, the process of generating the to-be-processed result after converting the input original image into the integer type is specifically expressed as follows: Wherein, the The original image is represented by a representation of the original image,The coordinates of the pixels are represented and,Representing the result to be processed. Further, the process of generating the primary blurred result of the smoothed image after the first gaussian blur processing is performed on the result to be processed specifically includes: constructing a Gaussian kernel K with a radius r, wherein the Gaussian kernel K is positioned at the position Weight value atExpressed as: The result to be processed Generating a primary blurring result of a smooth image after convolution operation with the gaussian kernel K based on a first gaussian blurring processAnd is expressed as: in the above, the result is blurred once Represented as pixelsThe gray value of the smoothed image at that point,Representing the gaussian kernel standard deviation, exp () represents the natural exponential function. Further, the convolution operation of the second gaussian blur processing is executed according to the absolute value of the difference between the to-be-processed result and the primary blur result, and the process of generating the secondary blur result specifically includes: Generating a residual error value by making a difference between the to-be-processed result and the primary fuzzy result Expressed as: for the residual value Taking absolute value to obtainExpressed as: performing convolution operation of the second Gaussian blur process to generate a secondary blur result Expressed as: In the above, the secondary blurring result Represented as pixelsAnd (5) gray values after secondary Gaussian blur. Further, the process of estimating the noise variance according to the secondary blurring result and generating the adaptive filter coefficient specifically includes: performing noise variance estimation processing according to the secondary blurring result to generate a first noise variance estimation value Expressed as: From the noise variance estimate Standard deviation from initial