CN104680483B - Image noise estimation method, video image denoising method and device - Google Patents

Abstract

The embodiment of the invention provides a noise estimation method of an image, a denoising method of a video image and a device, which are used for determining Laplacian operators in two directions of the image to be processed, wherein in a difference operator obtained by carrying out preset linear operation on the Laplacian operators in the two directions, the sum of numerical values of each row and the sum of numerical values of each column are zero, and the absolute value of the numerical value in the center is greater than the absolute value of the numerical value in a non-center position; smoothing the image to be processed by using a difference operator to obtain a smooth image; based on the first preset size, carrying out non-overlapping blocking on the smooth image to obtain each image block; and respectively determining the variances of each image block, determining a preset number of variances according to the sequence of variance values from small to large from each variance, and determining the weighted average of the preset number of variances as the estimated noise variance of the image to be processed. The invention solves the problem of inaccurate noise estimation of an image in the prior art, and relates to the field of image processing.

Description

Image noise estimation method, video image denoising method and device

technical field

The invention relates to the field of image processing, in particular to an image noise estimation method, a video image denoising method and a device.

Background technique

During the transmission of images or videos, they will be polluted by various noises, which will cause the resolution of the images or videos received by the receiver to be reduced compared with the original images or videos, which not only affects the visual effect, but also affects the image or video that needs to be obtained or recognized. Images or videos of moving targets affect the accuracy of acquisition or recognition. Therefore, in the prior art, it is necessary to estimate noise in an image or a video frame image, so as to prepare for subsequent denoising processing.

In the prior art, methods for estimating noise in images or video frame images mainly include the following steps:

Step 1, performing non-overlapping block processing on the image to be estimated or the video frame image;

Step 2, performing variance calculation on each image block obtained by block processing in step 1;

Step 3. From the variances calculated in step 2, select multiple values with smaller variances for noise estimation.

However, since there are edges in the target in the image or video frame image, and the target's edge affects the smoothness of the image, because the pixel value fluctuation of the smooth block can be considered to be caused by noise, therefore, by estimating the variance of the smooth block, it is An estimate of the noise variance in the original image can be obtained, that is, after the block processing in step 1, the image block without the target edge will be relatively smooth, and the calculated variance is very similar to the noise variance, while the image with the target edge The block is not smooth enough, resulting in a large difference between the calculated variance and the noise variance. For low-noise images, due to the small noise, the estimated noise variance of the edge block will be greater than the actual noise variance, resulting in overestimation of its noise. For high For noisy images, due to the large noise, the estimated noise of the edge block will be smaller than the actual noise variance, resulting in an underestimation of its noise, making the noise estimation inaccurate.

Contents of the invention

Embodiments of the present invention provide an image noise estimation method, a video image denoising method and device, to solve the problem of inaccurate image noise estimation in the prior art.

Based on the above problems, an image noise estimation method provided by an embodiment of the present invention includes:

Determining Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions satisfy the following conditions: preset the Laplacian operators in the two directions In the difference operator obtained by linear operation, the sum of the values of each row and the sum of the values of each column are both zero, and the absolute value of the value at the center is greater than the absolute value of the value at the non-center position;

smoothing the image to be processed by using the difference operator to obtain a smooth image;

Based on the first preset size, the smooth image is divided into non-overlapping blocks to obtain each image block;

Determine the variance of each image block respectively, and

From each variance, a preset number of variances are determined in ascending order of variance values, and a weighted average of the preset number of variances is determined as the estimated noise variance of the image to be processed.

A video image denoising method provided by an embodiment of the present invention includes:

Based on the above-mentioned noise estimation method of an image, determine the noise variance of the current frame image of the video to be processed, and determine the motion detection threshold according to the noise variance;

Based on the second preset size, the current frame image and the previous frame image are respectively divided into non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image;

For each image block of the current frame image, determine the similarity between the image block and the image block corresponding to the position of the image block in the previous frame image; and

Comparing the similarity with the motion detection threshold, and determining that the image block is a moving image block or a still image block according to the comparison result;

Denoising processing is performed on moving image blocks and still image blocks in the current frame image respectively.

An image noise estimation device provided by an embodiment of the present invention includes:

An operator determination module, configured to determine Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions satisfy the following conditions: the Laplacian operators in the two directions In the difference operator obtained by the preset linear operation of the Las operator, the sum of the values of each row and the sum of the values of each column are both zero, and the absolute value of the value at the center is greater than the absolute value of the value at the non-center position;

a smoothing processing module, configured to use the difference operator to perform smoothing processing on the image to be processed to obtain a smoothed image;

The first block module is used to perform non-overlapping block on the smooth image based on the first preset size to obtain each image block;

a variance determination module, configured to determine the variance of each image block respectively, and

A noise estimation module, configured to determine a preset number of variances from the variances in ascending order of variance values, and determine a weighted average of the preset number of variances as the estimated value of the image to be processed noise variance.

A video image denoising device provided in an embodiment of the present invention includes:

A threshold determination module, configured to determine the noise variance of the current frame image of the video to be processed through the above image noise estimation device, and determine the motion detection threshold according to the noise variance;

The second block module is used to divide the current frame image and the previous frame image into non-overlapping blocks based on a second preset size, so as to obtain each image block of the current frame image and the previous frame image Each image block of

A similarity determination module, configured to, for each image block of the current frame image, determine the similarity between the image block and the image block corresponding to the position of the image block in the previous frame image; and

A motion detection module, configured to compare the similarity with the motion detection threshold, and determine that the image block is a motion image block or a still image block according to the comparison result;

The denoising processing module is used to respectively perform denoising processing on the moving image block and the still image block in the current frame image.

The beneficial effects of the embodiments of the present invention include:

The image noise estimation method, video image denoising method and device provided by the embodiments of the present invention determine the Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions satisfy the following Condition: In the difference operator obtained by performing preset linear operations on Laplacian operators in two directions, the sum of the values in each row and the sum of the values in each column are both zero, and the absolute value of the value at the center is greater than that at the non- The absolute value of the numerical value of the center position; use the difference operator to smooth the image to be processed to obtain a smooth image; based on the first preset size, the smooth image is divided into non-overlapping blocks to obtain each image block; determine each image block respectively variance, and from each variance, determine a preset number of variances in ascending order of variance values, and determine the weighted average of the preset number of variances as the estimated noise variance of the image to be processed. Before the image to be processed is divided into blocks, the image is smoothed so that the target edge in the image becomes smooth, then the image in each divided image block becomes smooth, and the variance calculated by each image block can be relatively normal. Using It is more accurate to estimate the noise of the image with the variance;

Further, the noise variance of the current frame image of the video to be processed is determined using the above noise estimation method, and the motion detection threshold is determined according to the noise variance; based on the second preset size, the current frame image and the previous frame image are respectively non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image; for each image block of the current frame image, determine the difference between the image block and the image block corresponding to the image block position in the previous frame image The similarity between them; and the similarity is compared with the motion detection threshold, and according to the comparison result, it is determined that the image block is a moving image block or a still image block; respectively denoising the moving image block and the still image block in the current frame image deal with. In the above process of denoising the video image, noise estimation is performed on the image based on an accurate noise estimation method, and then denoising processing is performed, so that the denoising processing effect is better.

Description of drawings

FIG. 1 is a flowchart of an image noise estimation method provided by an embodiment of the present invention;

FIG. 2 is a flow chart of a video image denoising method provided by an embodiment of the present invention;

FIG. 3 is a flowchart of an image noise estimation method provided in Embodiment 1 of the present invention;

Figure 4a-Figure 4c is an example of the Laplacian operator template provided by Embodiment 1 of the present invention;

FIG. 5 is a flow chart of a video image denoising method provided in Embodiment 2 of the present invention;

FIG. 6 is a flowchart of a video image denoising method provided by Embodiment 3 of the present invention;

FIG. 7 is a schematic structural diagram of an image noise estimation device provided by an embodiment of the present invention;

FIG. 8 is a schematic structural diagram of a video image denoising device provided by an embodiment of the present invention.

detailed description

The embodiment of the present invention provides an image noise estimation method, a video image denoising method and device. The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only for illustration and explanation The present invention is not intended to limit the present invention. And in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

An embodiment of the present invention provides an image noise estimation method, as shown in FIG. 1 , including:

S101. Determine the Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions meet the following conditions: preset the Laplacian operators in the two directions In the difference operator obtained by linear operation, the sum of the values in each row and the sum of the values in each column are both zero, and the absolute value of the value at the center is greater than the absolute value of the value at the non-center position.

S102. Using the difference operator determined in S101 to perform smoothing processing on the image to be processed to obtain a smoothed image.

S103. Based on the first preset size, divide the smooth image obtained in S102 into non-overlapping blocks to obtain image blocks.

S104. Determine respectively the variance of each image block determined in S103.

S105. From the variances determined in S104, determine a preset number of variances in ascending order of variance values.

S106. Determine the weighted average of the preset number of variances determined in S105 as the estimated noise variance of the image to be processed.

Correspondingly, based on the above image noise estimation method, an embodiment of the present invention also provides a video image denoising method, as shown in FIG. 2 , including:

S201. Determine the noise variance of the current frame image of the video to be processed based on the above image noise estimation method, and determine a motion detection threshold according to the noise variance.

S202. Based on the second preset size, respectively divide the current frame image and the previous frame image into non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image.

S203. For each image block of the current frame image, determine the similarity between the image block and the image block corresponding to the position of the image block in the previous frame image.

S204. Compare the similarity determined in S203 with the motion detection threshold determined in S201, and determine that the image block is a moving image block or a still image block according to the comparison result.

S205. Perform denoising processing on moving image blocks and still image blocks in the current frame image respectively.

Further, the execution of step S201 and step S202-step S203 are not performed sequentially.

The method and related equipment provided by the present invention will be described in detail below with specific embodiments in conjunction with the accompanying drawings.

Example 1:

In Embodiment 1 of the present invention, an image noise estimation method is provided, as shown in FIG. 3 , which specifically includes the following steps:

S301. Determine the Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions meet the following conditions: preset the Laplacian operators in the two directions In the difference operator obtained by linear operation, the sum of the values in each row and the sum of the values in each column are both zero, and the absolute value of the value at the center is greater than the absolute value of the value at the non-center position.

Further, assuming that the determined Laplacian operators in two directions of the image to be processed perform the linear operation as in formula (1), the difference operator dL is obtained:

dL=aL ₁ -bL ₂ ; (1)

Among them, L ₁ and L ₂ respectively represent the Laplacian operators in the two determined directions; and the difference operator dL satisfies the sum of the values of each row and the sum of the values of each column are zero, and the absolute value of the value at the center Absolute value that is greater than a numeric value elsewhere.

The following example illustrates:

Assuming that a and b are 2 and 1 respectively, Figure 4a-Figure 4c are templates corresponding to L ₁ , L ₂ and dL respectively. As shown in Figure 4a-Figure 4c, in the template corresponding to dL, the sum of the values of each row is 0, the sum of the values of each column is 0, and the absolute value of the value in the middle is 4, which is greater than the value of the value in other positions Therefore, L ₁ and L ₂ can satisfy the conditions, and L ₁ and L ₂ can be determined as Laplacian operators in the two selected directions.

Further, in the prior art, the zero-crossing property of the Laplacian operator can be used to locate the target edge in the image, but in this embodiment, due to the nature of the difference operator, the difference operator can be used to smooth the image to be processed deal with.

S302. Use the difference operator determined in S301 to perform smoothing processing on the image to be processed by using the following formula:

$\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Wherein, (x, y) represents the pixel in the image to be processed; v(x, y) represents the gray value of the pixel (x, y) before smoothing the pixel (x, y); n(x, y) represents the gray value of the pixel point (x, y) after smoothing the pixel point (x, y); the value of l is the number of rows or columns of the Laplacian operator template minus 1 divided by 2.

Furthermore, the (2l+1)×(2l+1) image block with (x, y) as the center pixel in formula (2) corresponds to the Laplacian template, and the center pixel (x , y) after smoothing, the gray value n(x, y) is each value in the Laplacian template dL(x+i, y+j) corresponding to the position of the value in the image block Multiply the gray values of the pixels, and then add the obtained values to replace the gray value of the central pixel (x, y), which is the processed gray value n(x, y).

S303. Based on the first preset size, divide the smooth image obtained in S302 into non-overlapping blocks to obtain image blocks.

S304. Determine respectively the variance of each image block determined in S303.

S305. From the variances determined in S304, determine a preset number of variances in ascending order of variance values.

Preferably, in this step, the variances can be sorted from small to large, and the top 6% variances can be determined.

S306. Determine the weighted average of the preset number of variances determined in S305 as the estimated noise variance of the image to be processed.

In this step, the weights of the determined variances may be equal, that is, the average value of a preset number of variances is determined as the estimated noise variance of the image to be processed.

Example 2:

In Embodiment 2 of the present invention, a video image denoising method is provided, as shown in FIG. 5 , which specifically includes the following steps:

S501. Determine the noise variance of the current frame image of the video to be processed based on the above image noise estimation method, and determine a motion detection threshold according to the noise variance.

Further, a preset coefficient may be used to multiply the noise variance according to actual needs, and the product may be determined as the motion detection threshold.

Further, there is no strict execution sequence between this step and steps S502-S503, that is, as long as the motion detection threshold is determined before step S504.

S502. Based on the second preset size, respectively divide the current frame image and the previous frame image into non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image.

S503. For each image block of the current frame image, determine the similarity between the image block and the image block corresponding to the position of the image block in the previous frame image.

S504. Compare the similarity determined in S503 with the motion detection threshold determined in S501, and determine that the image block is a moving image block or a still image block according to the comparison result.

S505. Perform denoising processing on the moving image area and the still image area in the current frame image respectively.

S506. Determine the gradient value of each pixel in the current frame image.

Further, there is no strict sequence between this step and the execution of step S501-step S505, that is, as long as the gradient value of each pixel in the current frame image is determined before step S507 is executed.

Further, this step can also be modified to: determine the gradient value of each pixel in the moving image block in the current frame image.

S507. From each pixel in the moving image block, determine a pixel whose gradient value is greater than a preset gradient threshold.

Further, in this step, the pixel points whose gradient values are greater than the preset gradient threshold may be regarded as the pixels of the edge of the moving object.

S508. For each pixel in the determined pixels, the sum of the gray value obtained after denoising the pixel and the gray value before denoising the pixel in a preset ratio , which is further determined as the gray value of the pixel after denoising processing.

Further, for the pixels in the moving image block and the still image block in the current frame image as the edge of the non-moving object, the gray value of the pixel obtained after the denoising process can be directly used as the pixel point after the denoising process. The gray value of the pixel point, and for the pixel point as the edge of the moving target, it needs to be further processed according to this step. After further processing, the fine edge of the moving object in the image can be completely preserved, so that no smear phenomenon will appear in the processed image.

In this step, assuming that the pixel as the edge of the moving target has a gray value of Y before the denoising process, and a gray value of Y' after the denoising process, and the preset ratio is α, then the pixel is denoised. After that, the final gray value can be: Y'+αY.

Example 3:

In Embodiment 3 of the present invention, a video image denoising method is provided, as shown in FIG. 6 , which specifically includes the following steps:

S601. Determine the noise variance of the current frame image of the video to be processed based on the above image noise estimation method, and determine a motion detection threshold according to the noise variance.

Further, there is no strict execution sequence between this step and steps S602-S604, that is, as long as the motion detection threshold is determined before step S605.

S602. Based on the second preset size, respectively divide the current frame image and the previous frame image into non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image.

S603. For each image block of the current frame image and the previous frame image, determine an image area composed of s×s image blocks centered on the image block in the frame where the image block is located.

S604. For each image block of the current frame image, the following formula is used to determine the similarity D(i, j, k) between the image block and the image block corresponding to the position of the image block in the previous frame image:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Among them, (i, j) represents the center coordinates of the image block; the value of h is (s-1)/2; k and k-1 represent the frame number; (m, n) represents the image centered on the image block The coordinates of the image blocks other than the image block in the area relative to the image block; S _B (m, n, k) represents the grayscale of all pixels in the image block whose center coordinates are (m, n) in the kth frame The sum of values; S _B (m,n,k-1) represents the sum of the gray values of all pixels in the image block whose center coordinates are (m,n) in the k-1th frame.

Further, assuming that the value of s is 3, that is, for each image block of the current frame image and the previous frame image, determine the image block and the frame in which the image block is located, and a 3× An image area composed of three image blocks includes nine image blocks in the image area centered on the image block. Then in formula (3), calculate the sum of the gray values of all pixels in each of the nine image blocks of the current frame, and the gray values of all pixels in each of the nine image blocks in the previous frame and determine the square of the difference between the sum of the gray values of all pixels in the image blocks corresponding to each two positions in each of the nine image blocks in the current frame and the previous frame, and will get The squares of the 9 differences are summed, and the sum is taken as the similarity of the image block.

S605. When the similarity determined in S604 is greater than the motion detection threshold determined in S601, determine that the image block in the current frame is a moving image block relative to the image block corresponding to the position of the image block in the previous frame;

When the similarity determined in S604 is not greater than the motion detection threshold determined in S601, it is determined that the image block in the current frame is a still image block relative to the image block corresponding to the position of the image block in the previous frame.

Further, in this step, when D(i, j, k) is greater than the motion detection threshold, each pixel in the image block represented by (i, j) can be marked as 1, otherwise, marked as 0, then An image block composed of pixels marked as 1 represents a moving image block, and an image block composed of pixels marked 0 represents a still image block.

S606. Perform denoising processing on moving image blocks and still image blocks in the current frame image respectively.

In this step, the improved 3D-Rational filtering formula can be used to denoise the still image block in the current frame image, as shown in formula (4):

${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; \&Omega;</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\frac{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}^{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

For each pixel in the still image block, it is assumed that the gray value of the pixel before the denoising process is x ₀ , and the gray value of the pixel after the denoising process is y ₀ , t and t-1 are respectively Used to identify the frame number (does not represent the power value), Ω represents the index set of nine pixels adjacent to the pixel, i and j represent any pixel in Ω, x _i and x _j represent i respectively The gray value of the pixel and the j pixel, _{k t} , At both represent the time domain denoising adjustment factor.

In this step, the improved fast NLM method can be used to denoise the moving image block in the current frame image, as shown in formula (5):

$\mathrm{<span\; class="notranslate">\; NLv</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; I</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

in,

$\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

In formula (5), NLv(i) represents the gray value of pixel i after denoising processing on pixel i, pixel j is the pixel point adjacent to pixel i, and v(j) represents the gray value of pixel i Point j is the gray value of pixel j before the denoising process, w(i, j) means that the gray value of pixel j before denoising process is used to determine pixel i after denoising process When the gray value of , the weights used.

Equation (6) is the calculation formula of weight w(i,j), where ||N(i)-N(j)|| ² represents the Euclidean distance between pixel i and pixel j, and h is the attenuation Factor, ▽v(i) represents the gradient value of pixel i, ▽v(j) represents the gradient value of pixel j, σ _▽ represents another gradient threshold, Indicates the mean value of pixel i, Represents the average value of pixel point j, η1 and η2 represent the _first average threshold and the _second average threshold respectively, preferably, η1 and _η2 can be 0.9 and _1.2 respectively.

Based on the same inventive concept, an embodiment of the present invention also provides an image noise estimation device. Since the principle of the problem solved by these devices is similar to the aforementioned image noise estimation method, the implementation of the device can refer to the implementation of the aforementioned method , the repetitions will not be repeated.

An image noise estimation device provided by an embodiment of the present invention, as shown in FIG. 7 , includes the following modules:

The operator determination module 701 is configured to determine Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions satisfy the following conditions: the Laplacian operators in the two directions In the difference operator obtained by the preset linear operation of the Placian operator, the sum of the values of each row and the sum of the values of each column are both zero, and the absolute value of the value at the center is greater than the absolute value of the value at the non-center position;

A smoothing processing module 702, configured to use the difference operator to perform smoothing processing on the image to be processed to obtain a smoothed image;

The first block module 703 is configured to perform non-overlapping block on the smooth image based on a first preset size to obtain each image block;

A variance determining module 704, configured to determine the variance of each image block respectively, and

The noise estimation module 705 is configured to determine a preset number of variances from the variances in ascending order of variance values, and determine a weighted average of the preset number of variances as the estimated image to be processed noise variance.

Further, the smoothing processing module 702 is specifically configured to use the difference operator to perform smoothing processing on the image to be processed by using the following formula:

$\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}}{\overset{\mathrm{<span\; class="notranslate">\; l</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

Wherein, (x, y) represents the pixel in the image to be processed; v(x, y) represents the gray value of the pixel (x, y) before smoothing the pixel (x, y); n(x, y) represents the gray value of the pixel point (x, y) after smoothing the pixel point (x, y); the value of l is the number of rows or columns of the difference operator template minus 1 Divide by 2.

Based on the same inventive concept, the embodiment of the present invention also provides a video image denoising device based on the above image noise estimation device. Since the principle of the problem solved by these devices is similar to the aforementioned video image denoising method, the implementation of the device Reference may be made to the implementation of the aforementioned methods, and repeated descriptions will not be repeated here.

A video image denoising device provided in an embodiment of the present invention, as shown in FIG. 8 , includes the following modules:

The threshold determination module 801 is used to determine the noise variance of the current frame image of the video to be processed through the above image noise estimation device, and determine the motion detection threshold according to the noise variance;

The second block module 802 is configured to perform non-overlapping block on the current frame image and the previous frame image based on the second preset size, so as to obtain each image block of the current frame image and the previous frame each image block of the image;

A similarity determining module 803, configured to, for each image block of the current frame image, determine the similarity between the image block and the image block corresponding to the position of the image block in the previous frame image; and

A motion detection module 804, configured to compare the similarity with the motion detection threshold, and determine that the image block is a motion image block or a still image block according to the comparison result;

The denoising processing module 805 is configured to perform denoising processing on moving image blocks and still image blocks in the current frame image respectively.

Further, the above-mentioned video image denoising device also includes:

Gradient value determination module 806, configured to determine the gradient value of each pixel in the current frame image;

The pixel point determining module 807 is configured to, after the denoising processing module respectively performs denoising processing on the moving image block and the still image block in the current frame image, from each pixel in the moving image block, Determining pixels whose gradient values are greater than a preset gradient threshold;

The edge processing module 808 is configured to, for each pixel in the determined pixels, compare the gray value obtained after denoising the pixel with the preset ratio of the gray value before denoising the pixel The sum of the gray values is further determined as the gray value after the denoising process is performed on the pixel.

Further, the similarity determining module 803 is specifically configured to, for each image block of the current frame image and the previous frame image, determine the image block and the frame in which the image block is located, taking the image block as An image area composed of s×s image blocks in the center;

For each image block of the current frame, the following formula is used to determine the similarity D (i, j, k) between the image block and the image block corresponding to the position of the image block in the previous frame:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\overset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; [</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Among them, (i, j) represents the center coordinates of the image block; the value of h is (s-1)/2; k and k-1 represent the frame number; (m, n) represents the image centered on the image block The coordinates of the image blocks other than the image block in the area relative to the image block; S _B (m, n, k) represents the grayscale of all pixels in the image block whose center coordinates are (m, n) in the kth frame The sum of values; S _B (m,n,k-1) represents the sum of the gray values of all pixels in the image block whose center coordinates are (m,n) in the k-1th frame.

Further, the motion detection module 804 is specifically configured to determine that the image block in the current frame is a moving image relative to the image block corresponding to the image block position in the previous frame when the similarity is greater than the motion detection threshold block; when the similarity is not greater than the motion detection threshold, determine that the image block in the current frame is a still image block relative to the image block corresponding to the position of the image block in the previous frame.

The functions of the above-mentioned units may correspond to corresponding processing steps in the flow charts shown in FIG. 1 to FIG. 3 and FIG. 5-FIG.

The image noise estimation method, video image denoising method and device provided by the embodiments of the present invention determine the Laplacian operators in two directions of the image to be processed, wherein the Laplacian operators in the two directions satisfy the following Condition: In the difference operator obtained by performing preset linear operations on Laplacian operators in two directions, the sum of the values in each row and the sum of the values in each column are both zero, and the absolute value of the value at the center is greater than that at the non- The absolute value of the numerical value of the center position; use the difference operator to smooth the image to be processed to obtain a smooth image; based on the first preset size, the smooth image is divided into non-overlapping blocks to obtain each image block; determine each image block respectively variance, and from each variance, determine a preset number of variances in ascending order of variance values, and determine the weighted average of the preset number of variances as the estimated noise variance of the image to be processed. Before the image to be processed is divided into blocks, the image is smoothed so that the target edge in the image becomes smooth, then the image in each divided image block becomes smooth, and the variance calculated by each image block can be relatively normal. Using It is more accurate to estimate the noise of the image with the variance;

Further, the noise variance of the current frame image of the video to be processed is determined using the above noise estimation method, and the motion detection threshold is determined according to the noise variance; based on the second preset size, the current frame image and the previous frame image are respectively non-overlapping blocks to obtain each image block of the current frame image and each image block of the previous frame image; for each image block of the current frame image, determine the difference between the image block and the image block corresponding to the image block position in the previous frame image The similarity between them; and the similarity is compared with the motion detection threshold, and according to the comparison result, it is determined that the image block is a moving image block or a still image block; respectively denoising the moving image block and the still image block in the current frame image deal with. In the above process of denoising the video image, noise estimation is performed on the image based on an accurate noise estimation method, and then denoising processing is performed, so that the denoising processing effect is better.

Through the above description of the implementation manners, those skilled in the art can clearly understand that the embodiments of the present invention can be implemented by hardware, or by means of software plus a necessary general hardware platform. Based on this understanding, the technical solutions of the embodiments of the present invention can be embodied in the form of software products, which can be stored in a non-volatile storage medium (which can be CD-ROM, U disk, mobile hard disk, etc.), Several instructions are included to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute the methods described in various embodiments of the present invention.

Those skilled in the art can understand that the drawing is only a schematic diagram of a preferred embodiment, and the modules or processes in the drawing are not necessarily necessary for implementing the present invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description in the embodiment, or can be located in one or more devices different from the embodiment according to corresponding changes. The modules in the above embodiments can be combined into one module, and can also be further split into multiple sub-modules.

The serial numbers of the above embodiments of the present invention are for description only, and do not represent the advantages and disadvantages of the embodiments.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims (12)

1. A method of noise estimation of an image, comprising:

determining Laplacian operators of two directions of an image to be processed, wherein the Laplacian operators of the two directions meet the following conditions: in the difference operator obtained by performing preset linear operation on the laplacian operators in the two directions, the sum of numerical values of each row and the sum of numerical values of each column are all zero, the absolute value of the numerical value positioned in the center is greater than the absolute value of the numerical value positioned in a non-center position, and the preset linear operation formula is as follows: dL ═ aL₁-bL₂Which isIn, L₁And L₂Respectively representing Laplacian operators in two determined directions, dL represents the difference operator, a and b are respectively constants, and the number of rows and the number of columns of the Laplacian operators in the two determined directions are equal and are odd numbers;

smoothing the image to be processed by using the difference operator to obtain a smoothed image;

based on a first preset size, carrying out non-overlapping blocking on the smooth image to obtain each image block;

determining the variance of each image block, respectively, and

and determining a preset number of variances according to the sequence of variance values from small to large from each variance, and determining the weighted average value of the preset number of variances as the estimated noise variance of the image to be processed.

2. The method of claim 1, wherein the difference operator is used to smooth the image to be processed using the following formula:

$n(x,y)=\underset{j=-l}{\overset{l}{\&Sigma;}}\underset{i=-l}{\overset{l}{\&Sigma;}}dL(x+i,y+j)v(x+i,y+j);$

wherein, (x, y) represents a pixel point in the image to be processed; v (x, y) represents the gray value of the pixel point (x, y) before the smoothing processing is performed on the pixel point (x, y); n (x, y) represents the gray value of the pixel point (x, y) after the smoothing processing is performed on the pixel point (x, y); the value of l is the number of rows or columns of the Laplace operator template minus 1 divided by 2.

3. A video image denoising method is characterized by comprising the following steps:

the method for estimating noise based on the image according to any one of claims 1-2, determining the noise variance of the current frame image of the video to be processed, and determining the motion detection threshold according to the noise variance;

respectively carrying out non-overlapping blocking on the current frame image and the previous frame image based on a second preset size to obtain each image block of the current frame image and each image block of the previous frame image;

for each image block of the current frame image, determining the similarity between the image block and an image block corresponding to the position of the image block in the previous frame image; and are

Comparing the similarity with the motion detection threshold, and determining the image block to be a motion image block or a static image block according to the comparison result;

and denoising the moving image block and the static image block in the current frame image respectively.

4. The method of claim 3, further comprising:

determining the gradient value of each pixel point in the current frame image;

after the denoising processing is respectively performed on the moving image block and the static image block in the current frame image, the method further includes:

determining pixel points with gradient values larger than a preset gradient threshold value from all pixel points in the moving image blocks;

and aiming at each pixel point in the determined pixel points, further determining the sum of the gray value obtained after the pixel point is subjected to denoising processing and the gray value before the pixel point is subjected to denoising processing in a preset proportion as the gray value after the pixel point is subjected to denoising processing.

5. The method as claimed in claim 3, wherein determining, for each image block of the current frame image, a similarity between the image block and an image block corresponding to the position of the image block in the previous frame image comprises:

determining an image area formed by s multiplied by s image blocks taking the image block as a center in a frame where the image block is located aiming at each image block of the current frame image and the previous frame image;

for each image block of the current frame, determining the similarity D (i, j, k) between the image block and the image block corresponding to the position of the image block in the previous frame by adopting the following formula:

$D(i,j,k)=\underset{m=i-h}{\overset{i+h}{\&Sigma;}}\underset{n=j-h}{\overset{j+h}{\&Sigma;}}{\&lsqb;S_B(m,n,k)-S_B(m,n,k-1)\&rsqb;}^2$

wherein, (i, j) represents the center coordinates of the image block; h has a value of (s-1)/2; k and k-1 represent frame numbers; (m, n) denotes coordinates of image blocks other than the image block with respect to the image block in an image area centered on the image block; s_B(m, n, k) represents the sum of gray values of all pixel points in the image block with the center coordinate (m, n) in the kth frame; s_BAnd (m, n, k-1) represents the sum of gray values of all pixel points in the image block with the central coordinate of (m, n) in the (k-1) th frame.

6. The method according to claim 3, wherein comparing the similarity with the motion detection threshold and determining the image block as a moving image block or a stationary image block according to the comparison result specifically comprises:

when the similarity is larger than the motion detection threshold, determining that the image block of the current frame is a motion image block relative to the image block corresponding to the position of the image block in the previous frame;

and when the similarity is not greater than the motion detection threshold, determining that the image block of the current frame is a static image block relative to the image block corresponding to the position of the image block in the previous frame.

7. An apparatus for estimating noise of an image, comprising:

the operator determining module is used for determining laplacian operators in two directions of the image to be processed, wherein the laplacian operators in the two directions meet the following conditions: in the difference operator obtained by performing preset linear operation on the laplacian operators in the two directions, the sum of numerical values of each row and the sum of numerical values of each column are all zero, and the absolute value of the numerical value positioned in the center is greater than that positioned in the non-centerThe absolute value of the numerical value of the heart position is preset with a linear operation formula as follows: dL ═ aL₁-bL₂Wherein L is₁And L₂Respectively representing Laplacian operators in two determined directions, dL represents the difference operator, a and b are respectively constants, and the number of rows and the number of columns of the Laplacian operators in the two determined directions are equal and are odd numbers;

the smoothing processing module is used for smoothing the image to be processed by utilizing the difference operator to obtain a smooth image;

the first blocking module is used for carrying out non-overlapping blocking on the smooth image based on a first preset size to obtain each image block;

a variance determining module for determining the variance of each image block respectively and

and the noise estimation module is used for determining a preset number of variances from the variance values according to the sequence from small to large of the variance values, and determining the weighted average value of the preset number of variances as the estimated noise variance of the image to be processed.

8. The apparatus according to claim 7, wherein the smoothing module is specifically configured to perform smoothing on the image to be processed by using the difference operator according to the following formula:

$n(x,y)=\underset{j=-l}{\overset{l}{\&Sigma;}}\underset{i=-l}{\overset{l}{\&Sigma;}}dL(x+i,y+j)v(x+i,y+j);$

wherein, (x, y) represents a pixel point in the image to be processed; v (x, y) represents the gray value of the pixel point (x, y) before the smoothing processing is performed on the pixel point (x, y); n (x, y) represents the gray value of the pixel point (x, y) after the smoothing processing is performed on the pixel point (x, y); the value of l is the number of rows or columns of the Laplace operator template minus 1 divided by 2.

9. A video image denoising apparatus, comprising:

a threshold determination module, configured to determine, by the image noise estimation apparatus according to any one of claims 7 to 8, a noise variance of a current frame image of the video to be processed, and determine a motion detection threshold according to the noise variance;

the second blocking module is used for respectively carrying out non-overlapping blocking on the current frame image and the previous frame image based on a second preset size to obtain each image block of the current frame image and each image block of the previous frame image;

the similarity determining module is used for determining the similarity between each image block of the current frame image and the image block corresponding to the position of the image block in the previous frame image; and are

The motion detection module is used for comparing the similarity with the motion detection threshold and determining the image block to be a motion image block or a static image block according to the comparison result;

and the denoising processing module is used for respectively denoising the moving image block and the static image block in the current frame image.

10. The apparatus of claim 9, further comprising:

the gradient value determining module is used for determining the gradient value of each pixel point in the current frame image;

the pixel point determining module is used for determining pixel points with gradient values larger than a preset gradient threshold value from all pixel points in the moving image block after the denoising processing module denoises the moving image block and the static image block in the current frame image respectively;

and the edge processing module is used for further determining the sum of the gray value obtained after the pixel is subjected to denoising processing and the gray value before the pixel is subjected to denoising processing according to a preset proportion as the gray value after the pixel is subjected to denoising processing aiming at each pixel in the determined pixels.

11. The apparatus according to claim 9, wherein the similarity determining module is specifically configured to determine, for each image block of the current frame image and the previous frame image, an image area formed by s × s image blocks centered on the image block in a frame in which the image block is located;

for each image block of the current frame, determining the similarity D (i, j, k) between the image block and the image block corresponding to the position of the image block in the previous frame by adopting the following formula:

$D(i,j,k)=\underset{m=i-h}{\overset{i+h}{\&Sigma;}}\underset{n=j-h}{\overset{j+h}{\&Sigma;}}{\&lsqb;S_B(m,n,k)-S_B(m,n,k-1)\&rsqb;}^2$

wherein, (i, j) represents the center coordinates of the image block; h has a value of (s-1)/2; k and k-1 represent frame numbers; (m, n) denotes coordinates of image blocks other than the image block with respect to the image block in an image area centered on the image block; s_B(m, n, k) represents the sum of gray values of all pixel points in the image block with the center coordinate (m, n) in the kth frame; s_BAnd (m, n, k-1) represents the sum of gray values of all pixel points in the image block with the central coordinate of (m, n) in the (k-1) th frame.

12. The apparatus according to claim 9, wherein the motion detection module is specifically configured to determine that the image block of the current frame is a motion image block relative to an image block corresponding to the image block position in a previous frame when the similarity is greater than the motion detection threshold; and when the similarity is not greater than the motion detection threshold, determining that the image block of the current frame is a static image block relative to the image block corresponding to the position of the image block in the previous frame.