WO2016131370A1 - Wavelet denoising method and device - Google Patents

Abstract

A wavelet denoising method and device, comprising: downsampling an image, and respectively performing wavelet denoising on an image of each channel of each downsampled image to obtain a denoised image of each channel of each downsampled image; and synthesizing each obtained denoised image into a complete denoised image. The solution improves denoising efficiency.

Description

Method and device for implementing wavelet denoising Technical field

This document relates to, but is not limited to, image processing technology, and more particularly to a method and apparatus for implementing wavelet denoising.

Background technique

In the related art, the method for implementing wavelet denoising generally includes:

The image is subjected to dual extension; the image of the dual extension is wavelet transformed to obtain the first layer of wavelet coefficients; the noise variance of the image is calculated according to the high frequency part of the first layer of wavelet coefficients, and the noise variance of the image and the variance of the image are calculated. The wavelet coefficient of the first layer image is obtained by wavelet transforming the low frequency part of the i-th wavelet coefficient to obtain the (i+1)-th layer wavelet coefficient; and calculating the (i+1) according to the noise variance of the image and the low-frequency part of the i-th wavelet coefficient a layer image wavelet coefficient; wherein i is an integer greater than or equal to 1; performing inverse wavelet transform on the calculated image wavelet coefficients of each layer to obtain a denoised image.

In the related method of denoising, since wavelet denoising is only applicable to Gaussian distribution noise, when the noise completely conforms to the Gaussian distribution, the noise variance of the image is calculated according to the high frequency part of the first layer wavelet coefficient, and the actual The upper noise does not completely conform to the Gaussian distribution in the image, so the denoising effect is not good.

Summary of the invention

The following is an overview of the topics detailed in this document. This Summary is not intended to limit the scope of the claims.

The embodiment of the invention provides a method and device for implementing wavelet denoising, which can improve the denoising effect.

Embodiments of the present invention provide a method for implementing wavelet denoising, including:

Down-sampling the image, respectively performing wavelet denoising on the image of each channel of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling;

Each denoised image obtained is synthesized into a complete denoised image.

Optionally, the method further includes: performing grayscale processing on the complete denoised image.

Optionally, performing grayscale processing on the complete denoised image includes:

Performing high-pass filtering on the image of the Y channel of the complete denoised image after low-pass filtering;

Converting the high-pass filtered image into a binary image;

Determining that one pixel of the complete denoised image is a color caused by noise, and a pixel value corresponding to one pixel point of the complete denoised image in the binary image is 1, a decrease The pixel values of the U channel and the V channel of one pixel of the complete denoised image.

Optionally, converting the high-pass filtered image into a binary image includes:

Determining, that the pixel value of one pixel of the high-pass filtered image is greater than or equal to a first preset value, and the pixel value of the pixel corresponding to one pixel of the high-pass filtered image Set to 1;

Determining, that the pixel value of one pixel of the high-pass filtered image is smaller than the first preset value, and the pixel value of the pixel corresponding to one pixel of the high-pass filtered image Set to 0.

Optional,

When the average value of the pixel values of all the pixels of the high-pass filtered image is less than or equal to 60 divided by 255, the first preset value is 45 divided by 255;

When the average value is greater than 60 divided by 255, and less than or equal to 70 divided by 255, the first preset value is 40 divided by 255;

When the average value is greater than 70 divided by 255, and less than or equal to 80 divided by 255, the first preset value is 35 divided by 255;

When the average value is greater than 80 divided by 255, and less than or equal to 90 divided by 255, the first preset value is 30 divided by 255;

When the average value is greater than 90 divided by 255, the first preset value is 20 divided by 255.

Optionally, the determining, by the pixel of the complete denoised image, the color caused by the noise comprises:

Determining that a pixel of the complete denoised image satisfies abs(1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; wherein v is the complete go The pixel value of the V channel of one pixel of the noise image, u is the pixel value of the U channel of one pixel of the complete denoised image, t is the second preset value; abs() represents the absolute value.

Optionally, the reducing the pixel values of the U channel and the V channel of one pixel of the complete denoised image includes:

Calculating a ratio between a pixel value of a U channel of one pixel of the complete denoised image and a second preset value as a new pixel value of a U channel of one pixel of the complete denoised image, The ratio between the pixel value of the V channel of one pixel of the complete denoised image and the second preset value is used as a new pixel value of the V channel of one pixel of the complete denoised image.

Optionally, when the pixel value of the Y channel of one pixel of the complete denoised image is less than or equal to 30 divided by 255, the second preset value is 1.8;

When the pixel value of the Y channel of one pixel of the complete denoised image is greater than 30 divided by 255 and less than or equal to 60 divided by 255, the second preset value is 1.6;

When the pixel value of the Y channel of one pixel of the complete denoised image is greater than 60 divided by 255 and less than or equal to 90 divided by 255, the second preset value is 1.4;

The second preset value is 1.2 when the pixel value of the Y channel of one pixel of the complete denoised image is greater than 90 divided by 255 and less than or equal to 255 divided by 255.

Optionally, performing wavelet denoising on each channel of each channel of the downsampled image to obtain a denoised image of each channel of each image after downsampling includes:

Performing wavelet transform on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling;

Calculating the noise variance of the image of each channel of each image after down-sampling according to the high-frequency portion of the first layer wavelet coefficient of the image of each channel of each image after down-sampling, according to each image after down-sampling Calculating the first-layer image wavelet coefficients of the image of each channel of each image after down-sampling by calculating the noise variance of the image of each channel and the variance of the image of each channel of each image after down-sampling;

The low frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling Performing a wavelet transform to obtain an (i+1)th layer wavelet coefficient of an image of each channel of each image after downsampling; wherein i is an integer greater than or equal to 1;

Calculating the low frequency of the i-th wavelet coefficient of the image of each channel of each image after down-sampling according to the high-frequency portion of the (i+1)-th layer wavelet coefficient of the image of each channel of each image after down-sampling Partial noise variance, based on the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling and the i-th wavelet of the image of each channel of each image after down-sampling The variance of the low frequency portion of the coefficient calculates the (i+1)th layer image wavelet coefficient of the image of each channel of each image after downsampling;

The wavelet coefficients of each layer of the image of each channel of each channel of the downsampled image are respectively subjected to wavelet inverse transform to obtain a denoised image of each channel of each image after downsampling.

Optionally, the first layer of wavelet coefficients of each channel of each image after downsampling is divided by the high frequency portion of the (i+1)th layer wavelet coefficient to calculate each image of the downsampled image. The noise variance of the low frequency portion of the i-th wavelet coefficient of the image of each channel includes:

According to the formula

Calculating a noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images; median() indicates a median value;

Where σ _ni is the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each of the downsampled images, and y _ni+1 is each of each of the downsampled images The first layer wavelet coefficient of the image of the channel is divided by the pixel matrix of the high frequency portion of the (i+1)th layer wavelet coefficient.

Optionally, the noise variance of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling, and the i-th image of each channel of each image after down-sampling The variance of the low frequency portion of the layer wavelet coefficients is calculated. The (i+1)th layer image wavelet coefficients of the image of each channel of each image after downsampling include:

According to the formula

Calculating a variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images, according to a formula

Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each of the downsampled images;

Among them, according to the formula

Calculate T _ik ;

among them,

a variance of a kth pixel point of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of the downsampled image, N is a number of pixels of a neighborhood window, and y _ijk is the downsampling The pixel value of the jth pixel of the neighborhood window of the kth pixel of the low frequency portion of the i-th layer wavelet coefficient of each channel of each image after the image; j is an integer from 1 to N, k is 1 to an integer of the number of pixels of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of the downsampled image, w _i+1k is each channel of each of the downsampled images The pixel value of the kth pixel of the (i+1)th layer image wavelet coefficient, y _i+1k1 is the (i+1)th of the image of each channel of the downsampled image The real part of the pixel value of the kth pixel of the layer image wavelet coefficient, y _i+1k2 is the (i+1)th layer image wavelet coefficient of the image of each channel of each of the downsampled images The imaginary part of the pixel value of k pixels.

An embodiment of the present invention provides an apparatus for implementing wavelet denoising, including:

a downsampling module, set to downsample the image;

The denoising module is configured to perform wavelet denoising on each channel of each of the downsampled images to obtain a denoised image of each channel of each image after downsampling.

Optionally, the device further includes:

A grayscale processing module is configured to perform grayscale processing on the complete denoised image.

Optionally, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and the pixel value of the pixel corresponding to one pixel of the complete denoised image in the binary image is 1, reducing the U channel and V of one pixel of the complete denoised image The pixel value of the channel.

Optionally, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image, and determining that the pixel value of one pixel of the high-pass filtered image is greater than or equal to a first preset value, The pixel value of the pixel corresponding to one pixel of the high-pass filtered image is set to 1; and the pixel value of one pixel of the high-pass filtered image is determined to be smaller than a first preset value, the pixel value of the pixel corresponding to one pixel of the high-pass filtered image is set to 0; and one pixel of the complete denoised image is determined as a color caused by noise, and a pixel value of a pixel corresponding to one pixel of the complete denoised image in the binary image is 1, and a U channel of one pixel of the complete denoised image is reduced And the pixel value of the V channel.

Optionally, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image, and converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image satisfies abs ( 1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; where v is the pixel value of the V channel of one pixel of the complete denoised image; abs() Representing an absolute value, u is the pixel value of the U channel of one pixel of the complete denoised image, t is a second preset value, and one of the binary image and the complete denoised image The pixel value of the pixel corresponding to the pixel point is 1, and the pixel values of the U channel and the V channel of one pixel of the complete denoised image are reduced.

Optionally, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and the pixel value of the pixel corresponding to one pixel of the complete denoised image in the binary image is 1, and the pixel value of the U channel of one pixel of the complete denoised image is calculated. And a ratio between the second preset value and the new pixel value of the U channel of one pixel of the complete denoised image, and calculating a pixel value of the V channel of one pixel of the complete denoised image and The ratio between the second preset values is a new pixel value of the V channel of one pixel of the complete denoised image.

Optionally, the denoising module is set to:

Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each image of each image after down-sampling The high frequency portion of the first layer of wavelet coefficients of the image of the channel calculates the noise variance of the image of each channel of each image after downsampling, according to the noise variance and the amplitude of the image of each channel of each image after downsampling Calculating the variance of the image of each channel of each channel of each image after downsampling, the variance of the image of each channel of each image after sampling; for each image after downsampling The low frequency portion of the i-th layer wavelet coefficient of the image of each channel is subjected to wavelet transform to obtain the (i+1)th layer wavelet coefficient of the image of each channel of each image after down-sampling; wherein i is greater than or equal to An integer of 1; the ith layer of the image of each channel of each image after downsampling is calculated from the high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after downsampling The noise variance of the low-frequency portion of the wavelet coefficients, based on the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling and the image of each channel of each image after down-sampling The variance of the low frequency portion of the i-th wavelet coefficient is calculated as the (i+1)th layer image wavelet coefficient of the image of each channel of each image after down-sampling; the image of each channel of each image after down-sampling The wavelet coefficients of each layer of the image are respectively subjected to wavelet inverse transform to obtain a denoised image of each channel of each image after down-sampling, and each denoised image obtained is synthesized into a complete denoised image.

Optionally, the first layer of wavelet coefficients of each channel of each image after downsampling is divided by the high frequency portion of the (i+1)th layer wavelet coefficient to calculate each image of the downsampled image. The noise variance of the low frequency portion of the i-th wavelet coefficient of the image of each channel includes:

According to the formula

Calculating a noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images; median() indicates a median value;

Where σ _ni is the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each of the downsampled images, and y _ni+1 is each of each of the downsampled images The first layer wavelet coefficient of the image of the channel is divided by the pixel matrix of the high frequency portion of the (i+1)th layer wavelet coefficient.

Optionally, the noise variance of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling and the i-th layer of the image of each channel of each image after down-sampling The variance of the low frequency portion of the wavelet coefficients is calculated. The (i+1)th layer image wavelet coefficients of the image of each channel of each image after downsampling include:

According to the formula

Calculating a variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images, according to a formula

Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each of the downsampled images;

Among them, according to the formula

Calculate T _ik ;

among them,

a variance of a kth pixel point of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of the downsampled image, N is a number of pixels of a neighborhood window, and y _ijk is the downsampling The pixel value of the jth pixel of the neighborhood window of the kth pixel of the low frequency portion of the i-th layer wavelet coefficient of each channel of each image after the image; j is an integer from 1 to N, k is 1 to an integer of the number of pixels of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of the downsampled image, w _i+1k is each channel of each of the downsampled images The pixel value of the kth pixel of the (i+1)th layer image wavelet coefficient, y _i+1k1 is the (i+1)th of the image of each channel of the downsampled image The real part of the pixel value of the kth pixel of the layer image wavelet coefficient, y _i+1k2 is the (i+1)th layer image wavelet coefficient of the image of each channel of each of the downsampled images The imaginary part of the pixel value of k pixels.

The embodiment of the invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions are used to execute the above method.

Compared with the related art, the technical solution of the embodiment of the present invention includes: downsampling an image, respectively performing wavelet denoising on the image of each channel of each image after down-sampling to obtain each channel of each image after down-sampling Noise image; each denoised image obtained is synthesized into a complete denoised image. Through the scheme of the embodiment of the invention, the image is downsampled and then wavelet denoising is performed, and the noise of the image after downsampling is more consistent with the Gaussian distribution, thereby improving the denoising effect.

Further, calculating, according to the high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each of the downsampled images, the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each of the downsampled images The noise variance further improves the denoising effect.

Further, the complete denoising image is grayed out, and the noise points in the flat region are removed, so that the denoising effect is more optimized.

Other aspects will be apparent upon reading and understanding the drawings and detailed description.

BRIEF abstract

1 is a flowchart of a method for implementing wavelet denoising according to an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of an apparatus for implementing wavelet denoising according to an embodiment of the present invention.

Embodiments of the invention

In order to facilitate the understanding of those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings, and is not intended to limit the scope of the present invention. It should be noted that the embodiments in the present application and the various manners in the embodiments may be combined with each other without conflict.

Referring to FIG. 1, an embodiment of the present invention provides a method for implementing wavelet denoising, including:

Step 100: Perform down-sampling on the image, and perform wavelet denoising on the image of each channel of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling.

Optionally, in this step, performing wavelet denoising on each channel of each of the downsampled images to obtain a denoised image of each channel of each image after downsampling includes:

Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each image of each image after down-sampling The high frequency portion of the first layer of wavelet coefficients of the image of the channel calculates the noise variance of the image of each channel of each image after downsampling, according to the noise variance and the amplitude of the image of each channel of each image after downsampling The variance of the image of each channel of each image after sampling is calculated. The first layer image wavelet coefficients of the image of each channel of each image after downsampling are calculated; the image of each channel of each image after downsampling The low frequency portion of the i-th wavelet coefficient is subjected to wavelet transform to obtain the (i+1)th layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each channel of each image after down-sampling The high frequency portion of the (i+1)th layer wavelet coefficient of the image calculates the noise variance of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling, according to each image after down-sampling of The variance of the low-frequency portion of the i-th wavelet coefficient of the image of the channel and the variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel after down-sampling are calculated for each image after down-sampling The (i+1)th layer image wavelet coefficient of the image of each channel; each wavelet image of each layer of the image of each channel of the downsampled image is subjected to wavelet inverse transform to obtain each image after downsampling Denoising image for each channel.

Where i is an integer greater than or equal to 1. The maximum value of i can be set according to actual needs.

In this step, the image may be an image in the YUV format, and if the image is an image in the RGB format, a formula may be used.

Convert it to an image in YUV format. Then the image has three channels.

In this step, each image after downsampling should be a square image with a side length of 2 ⁿ and n being an integer greater than or equal to 1. The specific value of n is determined according to actual needs. The smaller n is, the higher the accuracy, but the larger the calculation.

In addition, if the side length of the image is not 2 ⁿ before downsampling, the image needs to be filled in advance so that the side length of the image is 2 ⁿ .

For example, the image size is 2304×4096, and after filling, it becomes 2560×4096, so that down sampling can be performed to obtain 512×512 images, and three channels of 512×512 images are separately subjected to wavelet denoising.

In this step, how to separately perform wavelet transform on the image of each channel of each image after down-sampling to obtain the first layer of wavelet coefficients is a well-known technique of those skilled in the art, and is not used to limit the protection scope of the present invention. Let me repeat.

In this step, before the wavelet transform, the image of each channel of each image after down-sampling may be subjected to dual extension, which is well known to those skilled in the art, and is not used to limit the protection of the present invention. The scope is not repeated here.

Wherein, calculating a noise variance of the image of each channel of each of the downsampled images according to the high frequency portion of the first layer of wavelet coefficients of the image of each channel of each image after downsampling includes:

According to the formula

A noise variance of an image of each channel of each of the downsampled images is calculated; median() indicates a median value.

Where σ _n0 is the noise variance of the image of each channel of each image after down-sampling, and y _n1 is the pixel of the high-frequency portion of the first layer wavelet system of the image of each channel of each image after down-sampling matrix.

Wherein, the image of each channel of each image after down-sampling is calculated according to the noise variance of the image of each channel of each image after down-sampling and the variance of the image of each channel of each image after down-sampling The first layer image wavelet coefficients include:

According to the formula

Calculate the variance of the image of each channel of each image after downsampling, according to the formula

Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each image after downsampling;

Among them, according to the formula

Calculate T _0k ;

among them,

For the variance of the kth pixel of the image of each channel of each image after downsampling, N is the number of pixels of the neighborhood window, and y _0jk is the image of each channel of each image after downsampling The pixel value of the jth pixel of the neighborhood window of k pixels; j is an integer from 1 to N, and k is an integer from 1 to the number of pixels of the image of each channel of each image after downsampling, w _1k is the pixel value of the kth pixel of the first layer image wavelet coefficient of the image of each channel of each image after downsampling, and y _1k1 is the image of each channel of each image after downsampling The real part of the pixel value of the kth pixel of the layer image wavelet coefficient, y _1k2 is the pixel value of the kth pixel of the first layer image wavelet coefficient of the image of each channel of each image after downsampling The imaginary part.

Wherein, how to perform wavelet transform on the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling to obtain (i+1) of the image of each channel of each image after down-sampling Layer wavelet coefficients are well known to those skilled in the art and are not intended to limit the scope of the present invention, and are not described herein again.

Wherein, the ith layer wavelet coefficient of the image of each channel of each image after downsampling is calculated according to the high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after downsampling The noise variance of the low frequency part includes:

According to the formula

Calculating a noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each image after down-sampling;

Where σ _ni is the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling, and y _ni+1 is the image of each channel of each image after down-sampling A pixel matrix of a high frequency portion of the (i+1)th layer wavelet coefficient.

Wherein, the noise variance of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling and the i-th layer wavelet of the image of each channel of each image after down-sampling The variance of the low frequency portion of the coefficient is calculated. The (i+1)th layer image wavelet coefficients of the image of each channel of each image after downsampling include:

According to the formula

Calculating the variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling, according to the formula

Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each image after downsampling;

Among them, according to the formula

Calculate T _ik ;

among them,

The variance of the kth pixel of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling, N is the number of pixels of the neighborhood window, and y _ijk is each of the downsampled The pixel value of the jth pixel of the neighborhood window of the kth pixel of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of the image; j is an integer from 1 to N, and k is 1 to downsampled The integer of the number of pixels of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after that, w _i+1k is the (i+1) of the image of each channel of each image after down-sampling The pixel value of the kth pixel of the layer image wavelet coefficient, y _i+1k1 is the kth pixel of the (i+1)th layer image wavelet coefficient of the image of each channel of each image after downsampling The real part of the pixel value, y _{i+1k2 ,} is the imaginary part of the pixel value of the kth pixel point of the (i+1)th layer image wavelet coefficient of the image of each channel of each image after downsampling.

Wherein, how to perform wavelet inverse transform on each layer of image wavelet coefficients of each channel image of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling, each of which will be obtained The denoising image synthesis of the complete denoising image is well known to those skilled in the art and is not intended to limit the scope of the present invention, and details are not described herein.

Step 101: Combine each of the obtained denoised images into a complete denoised image.

Through the scheme of the embodiment of the invention, the image is downsampled and then wavelet denoising is performed, and the noise of the image after downsampling is more consistent with the Gaussian distribution, thereby improving the denoising effect.

Further, the ith layer wavelet of the image of each channel of each image after downsampling is calculated according to the high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after downsampling The noise variance of the low frequency portion of the coefficient further improves the denoising effect.

Optionally, the method further includes:

Step 102: Perform grayscale processing on the complete denoised image. include:

Performing high-pass filtering on the image of the Y channel of the complete denoised image after low-pass filtering; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is color caused by noise, and The pixel value of the pixel corresponding to one pixel of the complete denoised image in the binary image is 1, reducing the pixel values of the U channel and the V channel of one pixel of the complete denoised image.

The high-pass filtering of the Y-channel image of the complete denoised image is performed by convolving the image of the Y-channel of the complete denoised image with the low-pass filter, and then performing the volume with the high-pass filter. product.

Among them, the low pass filter can be [0.125, 0.375, 0.125] ^T .

The high pass filter can be [0.125, 0, 0.375, 0, 0.375, 0, 0.125] ^T .

The converting the high-pass filtered image into a binary image includes:

Determining that the pixel value of one pixel of the high-pass filtered image is greater than or equal to the first preset value, and setting the pixel value of the pixel corresponding to one pixel of the high-pass filtered image to 1;

It is determined that the pixel value of one pixel of the high-pass filtered image is smaller than the first preset value, and the pixel value of the pixel corresponding to one pixel of the high-pass filtered image is set to zero.

Wherein, when the average value of the pixel values of all the pixels of the high-pass filtered image is less than or equal to 60 divided by 255, the first preset value is 45 divided by 255; when the average value is greater than 60 divided by 255, and less than or When the value is equal to 70 divided by 255, the first preset value is 40 divided by 255; when the average value is greater than 70 divided by 255, and less than or equal to 80 divided by 255, the first preset value is 35 divided by 255; When the value is greater than 80 divided by 255 and less than or equal to 90 divided by 255, the first preset value is 30 divided by 255; when the average value is greater than 90 divided by 255, the first preset value is 20 divided by 255.

Wherein, determining the color of a pixel of the complete denoised image as noise includes:

Determine that a pixel of the complete denoised image satisfies abs(1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; where v is a complete denoised image The pixel value of the V channel of the pixel, u is the pixel of the U channel of one pixel of the complete denoised image The value, t is the second preset value; abs() indicates the absolute value.

Where t can be taken as 25 divided by 255.

Wherein, reducing the pixel values of the U channel and the V channel of one pixel of the complete denoised image includes:

Calculating the ratio between the pixel value of the U channel of one pixel of the complete denoised image and the second preset value as a new pixel value of the U channel of one pixel of the complete denoised image, and calculating the complete denoising The ratio between the pixel value of the V channel of one pixel of the image and the second preset value is the new pixel value of the V channel of one pixel of the complete denoised image.

Wherein, when the pixel value of the Y channel of one pixel of the complete denoised image is less than or equal to 30 divided by 255, the second preset value is 1.8; when the pixel of the Y channel of one pixel of the complete denoised image When the value is greater than 30 divided by 255 and less than or equal to 60 divided by 255, the second preset value is 1.6; when the pixel value of the Y channel of one pixel of the complete denoised image is greater than 60 divided by 255 and less than or equal to 90 When divided by 255, the second preset value is 1.4; when the pixel value of the Y channel of one pixel of the complete denoised image is greater than 90 divided by 255 and less than or equal to 255 divided by 255, the second preset value is 1.2.

In this step, the complete denoising image is grayed out, and the noise points in the flat region are removed, so that the denoising effect is more optimized.

The embodiment of the invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and the computer executable instructions are used to execute the above method.

Referring to FIG. 2, an embodiment of the present invention further provides an apparatus for implementing wavelet denoising, including:

a downsampling module, set to downsample the image;

The denoising module is configured to perform wavelet denoising on each channel of each of the downsampled images to obtain a denoised image of each channel of each image after downsampling.

The device of the embodiment of the present invention further includes:

A grayscale processing module is configured to perform grayscale processing on the complete denoised image.

In the apparatus of the embodiment of the present invention, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and the pixel value of the pixel corresponding to one pixel of the complete denoised image in the binary image is 1, reducing the U channel and V of one pixel of the complete denoised image The pixel value of the channel.

In the apparatus of the embodiment of the present invention, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image, and determining that the pixel value of one pixel of the high-pass filtered image is greater than or equal to a first preset value, The pixel value of the pixel corresponding to one pixel of the high-pass filtered image is set to 1; determining that the pixel value of one pixel of the high-pass filtered image is smaller than the first preset value And setting a pixel value of the pixel corresponding to one pixel of the high-pass filtered image to 0; determining that one pixel of the complete denoised image is a color caused by noise, and A pixel value of a pixel corresponding to one pixel of the complete denoised image in the binary image is 1, and a pixel value of a U channel and a V channel of one pixel of the complete denoised image is reduced. .

In the apparatus of the embodiment of the present invention, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image, and converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image satisfies abs ( 1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; where v is the pixel value of the V channel of one pixel of the complete denoised image; abs() Representing an absolute value, u is the pixel value of the U channel of one pixel of the complete denoised image, t is a second preset value, and one of the binary image and the complete denoised image The pixel value of the pixel corresponding to the pixel point is 1, and the pixel values of the U channel and the V channel of one pixel of the complete denoised image are reduced.

In the apparatus of the embodiment of the present invention, the grayscale processing module is configured to:

Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and an image of the binary image with the complete denoised image Calculating a pixel value of a pixel corresponding to the prime point as 1, and calculating a ratio between a pixel value of the U channel of one pixel of the complete denoised image and a second preset value as one of the complete denoised images Calculating a new pixel value of the U channel of the pixel, calculating a ratio between a pixel value of the V channel of one pixel of the complete denoised image and the second preset value as the complete denoised image The new pixel value of the V channel of a pixel.

In the device of the embodiment of the invention, the denoising module is set to:

Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each image of each image after down-sampling The high frequency portion of the first layer of wavelet coefficients of the image of the channel calculates the noise variance of the image of each channel of each image after downsampling, according to the noise variance and the amplitude of the image of each channel of each image after downsampling The variance of the image of each channel of each image after sampling is calculated. The first layer image wavelet coefficients of the image of each channel of each image after downsampling are calculated; the image of each channel of each image after downsampling The low frequency portion of the i-th wavelet coefficient is subjected to wavelet transform to obtain the (i+1)th layer wavelet coefficient of the image of each channel of each image after down-sampling; wherein i is an integer greater than or equal to 1; The high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after sampling is calculated by calculating the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling Noise side The noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling and the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling The variance of the (i+1) layer image wavelet coefficients of the image of each channel of each image after downsampling; the wavelet coefficients of each layer of the image of each channel of each image after downsampling are performed separately The inverse wavelet transform is used to obtain the denoised image of each channel of each image after down-sampling, and each denoised image obtained is synthesized into a complete denoised image.

In the device of the embodiment of the invention, the denoising module is set to:

Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain the first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to the formula

Calculating a noise variance of an image of each channel of each of the downsampled images, according to a noise variance of an image of each channel of each image after downsampling and each channel of each image after downsampling The variance of the image calculates a first layer image wavelet coefficient of the image of each channel of each image after downsampling; and wavelets the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling Transforming to obtain the (i+1)th layer wavelet coefficient of the image of each channel of each image after downsampling; wherein i is an integer greater than or equal to 1;

Calculating the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling, according to the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling The variance of the noise and the variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling are calculated as the (i+1)th layer of the image of each channel of each image after down-sampling Image wavelet coefficient; respectively, wavelet transform is performed on each layer of image wavelet coefficients of each channel image of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling, and each obtained will be obtained Denoising images are combined to form a complete denoised image.

In the device of the embodiment of the invention, the denoising module is set to:

Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each image of each image after down-sampling The high frequency portion of the first layer of wavelet coefficients of the image of the channel calculates the noise variance of the image of each channel of each image after downsampling, according to the formula

Calculate the variance of the image of each channel of each image after downsampling, according to the formula

Calculating a first layer image wavelet coefficient of an image of each channel of each of the downsampled images; wherein, according to a formula

Calculating T _0k ;; performing wavelet transform on the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling to obtain the image of each channel of each image after down-sampling (i+ 1) layer wavelet coefficients; wherein i is an integer greater than or equal to 1; the downsampled frequency is calculated according to the high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after downsampling The noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image, according to the formula

Calculating the variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling, according to the formula

Calculating the (i+1)th layer image wavelet coefficient of the image of each channel of each image after downsampling; wherein, according to the formula

Calculating T _ik ; performing inverse wavelet transform on the wavelet coefficients of each layer of the image of each channel of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling, each of which will be obtained The denoised image is combined to form a complete denoised image.

One of ordinary skill in the art will appreciate that all or a portion of the above steps may be performed by a program to instruct related hardware, such as a processor, which may be stored in a computer readable storage medium, such as a read only memory, disk or optical disk. Wait. Alternatively, all or part of the steps of the above embodiments may also be implemented using one or more integrated circuits. Correspondingly, each module/unit in the above embodiment may be implemented in the form of hardware, for example, by implementing an integrated circuit to implement its corresponding function, or may be implemented in the form of a software function module, for example, executing a program stored in the memory by a processor. / instruction to achieve its corresponding function. The invention is not limited to any specific form of combination of hardware and software.

It should be noted that the above-mentioned embodiments are only for the convenience of those skilled in the art, and are not intended to limit the scope of the present invention, and those skilled in the art without departing from the inventive concept of the present invention. Any obvious substitutions and improvements made to the invention are within the scope of the invention.

Industrial applicability

The above technical solution optimizes the denoising effect.

Claims (21)

1. A method for implementing wavelet denoising includes:

   Down-sampling the image, respectively performing wavelet denoising on the image of each channel of each image after down-sampling to obtain a denoised image of each channel of each image after down-sampling;

   Each denoised image obtained is synthesized into a complete denoised image.
2. The method of claim 1 further comprising:

   The complete denoised image is grayed out.
3. The method of claim 2 wherein said graying out the complete denoised image comprises:

   Performing high-pass filtering on the image of the Y channel of the complete denoised image after low-pass filtering;

   Converting the high-pass filtered image into a binary image;

   Determining that one pixel of the complete denoised image is a color caused by noise, and a pixel value corresponding to one pixel point of the complete denoised image in the binary image is 1, a decrease The pixel values of the U channel and the V channel of one pixel of the complete denoised image.
4. The method of claim 3 wherein said converting said high pass filtered image to a binary image comprises:

   Determining, that the pixel value of one pixel of the high-pass filtered image is greater than or equal to a first preset value, and the pixel value of the pixel corresponding to one pixel of the high-pass filtered image Set to 1;

   Determining, that the pixel value of one pixel of the high-pass filtered image is smaller than the first preset value, and the pixel value of the pixel corresponding to one pixel of the high-pass filtered image Set to 0.
5. The method of claim 4, wherein

   When the average value of the pixel values of all the pixels of the high-pass filtered image is less than or equal to 60 divided by 255, the first preset value is 45 divided by 255;

   When the average value is greater than 60 divided by 255 and less than or equal to 70 divided by 255, the first A preset value is 40 divided by 255;

   When the average value is greater than 70 divided by 255, and less than or equal to 80 divided by 255, the first preset value is 35 divided by 255;

   When the average value is greater than 80 divided by 255, and less than or equal to 90 divided by 255, the first preset value is 30 divided by 255;

   When the average value is greater than 90 divided by 255, the first preset value is 20 divided by 255.
6. The method according to claim 3, wherein said determining a color of a pixel of the complete denoised image as noise comprises:

   Determining that a pixel of the complete denoised image satisfies abs(1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; wherein v is the complete go The pixel value of the V channel of one pixel of the noise image, u is the pixel value of the U channel of one pixel of the complete denoised image, t is the second preset value; abs() represents the absolute value.
7. The method of claim 3, wherein the reducing the pixel values of the U channel and the V channel of one pixel of the complete denoised image comprises:

   Calculating a ratio between a pixel value of a U channel of one pixel of the complete denoised image and a second preset value as a new pixel value of a U channel of one pixel of the complete denoised image, The ratio between the pixel value of the V channel of one pixel of the complete denoised image and the second preset value is used as a new pixel value of the V channel of one pixel of the complete denoised image.
8. The method of claim 7 wherein

   When the pixel value of the Y channel of one pixel of the complete denoised image is less than or equal to 30 divided by 255, the second preset value is 1.8;

   When the pixel value of the Y channel of one pixel of the complete denoised image is greater than 30 divided by 255 and less than or equal to 60 divided by 255, the second preset value is 1.6;

   When the pixel value of the Y channel of one pixel of the complete denoised image is greater than 60 divided by 255 and less than or equal to 90 divided by 255, the second preset value is 1.4;

   The second preset value is 1.2 when the pixel value of the Y channel of one pixel of the complete denoised image is greater than 90 divided by 255 and less than or equal to 255 divided by 255.
9. The method according to claim 1 or 2, wherein said image of each channel of each of the downsampled images is subjected to wavelet denoising to obtain a denoised image of each channel of each image after downsampling include:

   Performing wavelet transform on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling;

   Calculating the noise variance of the image of each channel of each image after down-sampling according to the high-frequency portion of the first layer wavelet coefficient of the image of each channel of each image after down-sampling, according to each image after down-sampling Calculating the first-layer image wavelet coefficients of the image of each channel of each image after down-sampling by calculating the noise variance of the image of each channel and the variance of the image of each channel of each image after down-sampling;

   Performing wavelet transform on the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling to obtain the (i+1)-th layer wavelet coefficient of the image of each channel of each image after down-sampling Where i is an integer greater than or equal to 1;

   Calculating the low frequency of the i-th wavelet coefficient of the image of each channel of each image after down-sampling according to the high-frequency portion of the (i+1)-th layer wavelet coefficient of the image of each channel of each image after down-sampling Partial noise variance, based on the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling and the i-th wavelet of the image of each channel of each image after down-sampling The variance of the low frequency portion of the coefficient calculates the (i+1)th layer image wavelet coefficient of the image of each channel of each image after downsampling;

   The wavelet coefficients of each layer of the image of each channel of each channel of the downsampled image are respectively subjected to wavelet inverse transform to obtain a denoised image of each channel of each image after downsampling.
10. The method of claim 9, wherein the first layer wavelet coefficients of the image of each channel of each image after downsampling are divided by the high frequency portion of the (i+1)th layer wavelet coefficient The noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after sampling includes:

    According to the formula

    

    Calculating a noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images; median() indicates a median value;

    Where σ _ni is the noise variance of the low frequency portion of the i-th wavelet coefficient of the image of each channel of each of the downsampled images, and y _ni+1 is each of the downsampled images The first layer wavelet coefficient of the image of the channel is divided by the pixel matrix of the high frequency portion of the (i+1)th layer wavelet coefficient.
11. The method according to claim 9, wherein said noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each image after down-sampling, and each image of each image after down-sampling The variance of the low-frequency portion of the i-th wavelet coefficient of the image of the channel is calculated. The (i+1)-th layer image wavelet coefficients of the image of each channel of each image after down-sampling include:

    According to the formula

    

    Calculating a variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images, according to a formula

    

    Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each of the downsampled images;

    Among them, according to the formula

    

    Calculate T _ik ;

    among them,

    

    a variance of a kth pixel point of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of the downsampled image, N is a number of pixels of a neighborhood window, and y _ijk is the downsampling The pixel value of the jth pixel of the neighborhood window of the kth pixel of the low frequency portion of the i-th layer wavelet coefficient of each channel of each image after the image; j is an integer from 1 to N, k is 1 to an integer of the number of pixels of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of the downsampled image, w _i+1k is each channel of each of the downsampled images The pixel value of the kth pixel of the (i+1)th layer image wavelet coefficient, y _i+1k1 is the (i+1)th of the image of each channel of the downsampled image The real part of the pixel value of the kth pixel of the layer image wavelet coefficient, y _i+1k2 is the (i+1)th layer image wavelet coefficient of the image of each channel of each of the downsampled images The imaginary part of the pixel value of k pixels.
12. A device for implementing wavelet denoising, comprising:

    a downsampling module, set to downsample the image;

    The denoising module is configured to perform wavelet denoising on each channel of each of the downsampled images to obtain a denoised image of each channel of each image after downsampling.
13. The apparatus of claim 12, further comprising:

    A grayscale processing module is configured to perform grayscale processing on the complete denoised image.
14. The apparatus of claim 13 wherein said grayscale processing module is configured to:

    Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and the pixel value of the pixel corresponding to one pixel of the complete denoised image in the binary image is 1, reducing the U channel and V of one pixel of the complete denoised image The pixel value of the channel.
15. The apparatus of claim 13 wherein said grayscale processing module is configured to:

    Performing high-pass filtering on the image of the Y channel of the complete denoised image, and determining that the pixel value of one pixel of the high-pass filtered image is greater than or equal to a first preset value, The pixel value of the pixel corresponding to one pixel of the high-pass filtered image is set to 1; determining that the pixel value of one pixel of the high-pass filtered image is smaller than the first preset value And setting a pixel value of the pixel corresponding to one pixel of the high-pass filtered image to 0; determining that one pixel of the complete denoised image is a color caused by noise, and A pixel value of a pixel corresponding to one pixel of the complete denoised image in the binary image is 1, and a pixel value of a U channel and a V channel of one pixel of the complete denoised image is reduced. .
16. The apparatus of claim 13 wherein said grayscale processing module is configured to:

    Performing high-pass filtering on the image of the Y channel of the complete denoised image, and converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image satisfies abs ( 1.1398v)<t and abs(0.3946u+0.5806v)<t and abs(2.0321u)<t; where v is the pixel value of the V channel of one pixel of the complete denoised image; abs() Representing an absolute value, u is the pixel value of the U channel of one pixel of the complete denoised image, t is a second preset value, and one of the binary image and the complete denoised image The pixel value of the pixel corresponding to the pixel point is 1, and the pixel values of the U channel and the V channel of one pixel of the complete denoised image are reduced.
17. The apparatus of claim 13 wherein said grayscale processing module is configured to:

    Performing high-pass filtering on the image of the Y channel of the complete denoised image; converting the high-pass filtered image into a binary image; determining that one pixel of the complete denoised image is caused by noise Color, and an image of the binary image with the complete denoised image Calculating a pixel value of a pixel corresponding to the prime point as 1, and calculating a ratio between a pixel value of the U channel of one pixel of the complete denoised image and a second preset value as one of the complete denoised images Calculating a new pixel value of the U channel of the pixel, calculating a ratio between a pixel value of the V channel of one pixel of the complete denoised image and the second preset value as the complete denoised image The new pixel value of the V channel of a pixel.
18. The apparatus according to claim 12 or 13, wherein said denoising module is configured to:

    Wavelet transform is respectively performed on the image of each channel of each image after down-sampling to obtain a first layer wavelet coefficient of the image of each channel of each image after down-sampling; according to each image of each image after down-sampling The high frequency portion of the first layer of wavelet coefficients of the image of the channel calculates the noise variance of the image of each channel of each image after downsampling, according to the noise variance and the amplitude of the image of each channel of each image after downsampling The variance of the image of each channel of each image after sampling is calculated. The first layer image wavelet coefficients of the image of each channel of each image after downsampling are calculated; the image of each channel of each image after downsampling The low frequency portion of the i-th wavelet coefficient is subjected to wavelet transform to obtain the (i+1)th layer wavelet coefficient of the image of each channel of each image after down-sampling; wherein i is an integer greater than or equal to 1; The high frequency portion of the (i+1)th layer wavelet coefficient of the image of each channel of each image after sampling is calculated by calculating the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of each image after down-sampling Noise side The noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling and the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling The variance of the (i+1) layer image wavelet coefficients of the image of each channel of each image after downsampling; the wavelet coefficients of each layer of the image of each channel of each image after downsampling are performed separately The inverse wavelet transform is used to obtain the denoised image of each channel of each image after down-sampling, and each denoised image obtained is synthesized into a complete denoised image.
19. The apparatus according to claim 18, wherein said denoising module is configured to realize a high frequency of (i+1)th layer wavelet coefficient of an image of each channel of each image after down-sampling by: Partially calculating the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling:

    According to the formula

    

    Calculating a noise variance of a low frequency portion of an i-th layer wavelet coefficient of each channel of each downsampled image; median() means taking a median value;

    Where σ _ni is the noise variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each of the downsampled images, and y _ni+1 is each of each of the downsampled images The first layer wavelet coefficient of the image of the channel is divided by the pixel matrix of the high frequency portion of the (i+1)th layer wavelet coefficient.
20. The apparatus according to claim 18, wherein said denoising module is configured to realize a noise variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each image after down-sampling by: The variance of the low-frequency portion of the i-th wavelet coefficient of the image of each channel of each image after down-sampling is calculated as the (i+1)-th layer image wavelet coefficient of the image of each channel of each image after down-sampling:

    According to the formula

    

    Calculating a variance of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of each of the downsampled images, according to a formula

    

    Calculating an (i+1)th layer image wavelet coefficient of an image of each channel of each of the downsampled images;

    Among them, according to the formula

    

    Calculate T _ik ;

    among them,

    

    a variance of a kth pixel point of a low frequency portion of an i-th layer wavelet coefficient of an image of each channel of the downsampled image, N is a number of pixels of a neighborhood window, and y _ijk is the downsampling The pixel value of the jth pixel of the neighborhood window of the kth pixel of the low frequency portion of the i-th layer wavelet coefficient of each channel of each image after the image; j is an integer from 1 to N, k is 1 to an integer of the number of pixels of the low frequency portion of the i-th layer wavelet coefficient of the image of each channel of the downsampled image, w _i+1k is each channel of each of the downsampled images The pixel value of the kth pixel of the (i+1)th layer image wavelet coefficient, y _i+1k1 is the (i+1)th of the image of each channel of the downsampled image The real part of the pixel value of the kth pixel of the layer image wavelet coefficient, y _i+1k2 is the (i+1)th layer image wavelet coefficient of the image of each channel of each of the downsampled images The imaginary part of the pixel value of k pixels.
21. A computer storage medium having stored therein computer executable instructions for performing the method of any one of claims 1-11.

Images