WO2017036386A1 - Video denoising method and device, terminal and storage medium - Google Patents

Abstract

Disclosed in embodiments of the present invention is a video denoising method, comprising: detecting an input image of a target video to obtain the mosquito noise probability of each pixel in the input image; low-pass filtering the input image to obtain the pixel value corresponding to each filtered pixel; and on the basis of the mosquito noise probability, performing weighted processing on the pixel value after filtering and the pixel value of the input image to obtain the pixel value of an output image, and outputting the output image. The embodiments of the present invention also disclose a video denoising device, a terminal and a storage medium.

Description

Video denoising method and device, terminal and storage medium Technical field

The present invention relates to the field of image processing, and in particular, to a video denoising method and apparatus, a terminal, and a storage medium.

Background technique

In the process of video encoding and decoding, the original video data needs to be compressed. The commonly used compression coding algorithm usually has a coefficient quantization process, and the irreversibility of the quantization process causes a large amount of mosquito noise in the final decoded video data. Most of the noise is around the edge of the font or object, which causes the video quality to drop. At the same time, the video picture with mosquito noise makes the viewer feel "dirty" and affects the viewer's visual experience. In order to solve the above problem, the decoded video is subjected to a mosquito denoising operation, so that the final video looks clean and enhances the viewer's visual experience and video quality.

At present, the common method for removing mosquito noise is to use low-pass filtering, but the commonly used low-pass filter has the following problems: 1. Filtering the entire image; 2. The strength of the low-pass filter cannot be flexibly adjusted, resulting in high The loss of frequency detail components causes the video to be blurred.

Therefore, there is no suitable method for removing mosquito noise in video in the prior art.

Summary of the invention

In view of this, embodiments of the present invention are directed to a video denoising method and apparatus, a terminal, and a storage medium, so as to remove the mosquito noise existing in the video while preserving the details in the video and improving the viewer's vision. Feeling and video quality.

To achieve the above objective, the technical solution of the embodiment of the present invention is implemented as follows:

In a first aspect, an embodiment of the present invention provides a video denoising method, including: detecting a target view a frequency input image, obtaining a mosquito noise probability of each pixel in the input image; performing low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; a noise probability, weighting the filtered pixel value and the pixel value of the input image to obtain a pixel value of the output image, and outputting the output image.

In an embodiment of the present invention, the detecting an input image of the target video, obtaining a mosquito noise probability of each pixel in the input image, comprising: detecting an input image of the target video, obtaining the input image a pixel type of each of the pixels; based on the pixel type of each of the pixels, performing a mosquito noise probability estimation on each of the pixels in the input image to obtain a mosquito noise probability of each of the pixels.

In an embodiment of the present invention, the detecting an input image of the target video to obtain a pixel type of each pixel in the input image includes: performing gradient detection on the input image to obtain each of the input images a gradient value of one pixel; performing local edge detection on the input image to obtain an edge information value of each pixel of the input image; determining each of the input images based on the gradient value and the edge information value The pixel type of the pixel.

In an embodiment of the present invention, determining, according to the gradient value and the edge information value, a pixel type of each pixel of the input image, including: when the gradient value of the ith pixel is greater than When the product of the edge information value of the ith pixel is equal to the first preset value, the pixel type of the ith pixel is determined as an edge pixel, where i is a positive integer; The gradient value of the pixel is smaller than the product of the edge information value of the ith pixel and the first preset value, and is greater than or equal to the edge information value of the ith pixel and the second preset value. Determining the pixel type of the ith pixel as a detail pixel, wherein the first preset value is different from the second preset value; when the gradient value of the ith pixel When the product of the edge information value of the ith pixel and the first preset value is smaller than the product of the edge information value of the ith pixel and the second preset value, The pixel type of the i-th pixel is determined to be a flat pixel.

In an embodiment of the present invention, the low-pass filtering the input image to obtain the filtered pixel value corresponding to each pixel includes: based on the edge information value of each pixel The pixel values of the input image are subjected to low-pass bilateral filtering to obtain the filtered pixel values corresponding to each pixel.

In an embodiment of the present invention, the mosquito noise probability estimation is performed on each pixel in the input image based on a pixel type of each pixel, and mosquito noise of each pixel is obtained. The probability includes: counting the pixel types of the ith pixel and the M neighboring pixels, wherein the neighboring pixels are pixels around the ith pixel; determining the ith pixel based on a statistical result Mosquito noise probability.

In an embodiment of the present invention, determining the mosquito noise probability of the ith pixel based on a statistical result, comprising: determining, according to the ith pixel and the M pixel, a flat pixel The number of pixels is a ratio of the number of pixels of the pixel type to the detail pixel to the sum of the number of pixels of the pixel type of the edge pixel, and the mosquito noise probability of the ith pixel is determined.

In a second aspect, an embodiment of the present invention provides a video denoising apparatus, including: a detecting unit, a filtering unit, and an output unit; wherein the detecting unit is configured to detect an input image of a target video, and obtain each of the input images. a mosquito-like noise probability of one pixel; the filtering unit configured to perform low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; the output unit is configured to be based on the Mosquito noise probability, weighting the filtered pixel value and the pixel value of the input image to obtain a pixel value of the output image, and outputting the output image.

In an embodiment of the present invention, the detecting unit includes: a pixel type detecting unit configured to detect an input image of the target video, obtain a pixel type of each pixel in the input image; and a noise probability estimating unit And configuring, according to the pixel type of each of the pixels, performing a mosquito noise probability estimation on each of the pixels in the input image to obtain a mosquito noise probability of each pixel.

In an embodiment of the present invention, the pixel type detecting unit includes: a gradient detecting unit configured to perform gradient detection on the input image to obtain a gradient value of each pixel of the input image; and an edge detecting unit Configuring to perform local edge detection on the input image to obtain an edge information value of each pixel of the input image; a pixel type determining unit configured to determine the input based on the gradient value and the edge information value The pixel type of each pixel of the image.

In an embodiment of the present invention, the pixel type determining unit is configured to: when the gradient value of the ith pixel is greater than or equal to the edge information value of the ith pixel and the first preset value In the case of product, the pixel type of the ith pixel is determined as an edge pixel, where i is a positive integer; and is further configured to: when the gradient value of the ith pixel is smaller than the ith pixel When the product of the edge information value and the first preset value is greater than or equal to the product of the edge information value of the ith pixel and the second preset value, determining the pixel type of the ith pixel as the detail a pixel, wherein the first preset value is different from the second preset value; and configured to: when the gradient value of the ith pixel is smaller than the edge information value of the ith pixel When the product of the first preset value is smaller than the product of the edge information value of the ith pixel and the second preset value, the pixel type of the ith pixel is determined to be a flat pixel .

In an embodiment of the present invention, the filtering unit is configured to perform low-pass bilateral filtering based on edge information values of each pixel and pixel values of the input image to obtain corresponding to each pixel. Filtered pixel value.

In an embodiment of the present invention, the noise probability estimation unit includes: a pixel type statistical unit configured to perform statistics on pixel types of the ith pixel and the M neighborhood pixels, wherein the neighborhood pixel a pixel around the ith pixel; a probability calculation unit configured to determine a mosquito noise probability of the ith pixel based on a statistical result.

In an embodiment of the present invention, the probability calculation unit is configured to: based on the ith pixel and the M pixels, a pixel type of which is a flat pixel, and the pixel type is The ratio of the number of pixels of the detail pixel to the sum of the number of pixels of the pixel type of the edge pixel determines the mosquito noise probability of the ith pixel.

In a third aspect, an embodiment of the present invention provides a computer storage medium, where the computer storage medium stores computer executable instructions, and the computer executable instructions are used to perform the video denoising method provided by the first aspect of the present invention.

In a fourth aspect, an embodiment of the present invention provides a terminal, where the terminal includes:

a storage medium configured to store computer executable instructions;

a processor configured to detect an input image of the target video, obtain a mosquito noise probability of each pixel in the input image, and perform low-pass filtering on the input image to obtain a filtered pixel corresponding to each pixel And performing weighting processing on the filtered pixel value and the pixel value of the input image based on the mosquito noise probability, obtaining a pixel value of the output image, and outputting the output image.

Embodiments of the present invention provide a video denoising method and apparatus, a terminal, and a storage medium. The device detects an input image of a target video, obtains a mosquito noise probability of each pixel in the input image, and simultaneously performs a low pass on the input image. Filtering, obtaining a filtered pixel value corresponding to each pixel; then, based on the mosquito noise probability, weighting the filtered pixel value and the pixel value of the input image to obtain a pixel value of the output image, and outputting the output image. That is to say, for each frame of the target video, only pixels with a relatively high mosquito noise probability are filtered, and other pixels are not processed, so that not only the mosquito noise existing in the video but also the mosquito noise can be removed. The details in the video are preserved to enhance the viewer's visual experience and video quality.

DRAWINGS

1 is a schematic flowchart of a video denoising method according to an embodiment of the present invention;

2 is a schematic flow chart of a method for obtaining a mosquito noise probability of each pixel according to an embodiment of the present invention;

3 is a schematic diagram of a 3×3 template according to an embodiment of the present invention;

4 is a schematic diagram of an 11×11 template according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a 5×5 template according to an embodiment of the present invention; FIG.

6 is a schematic diagram of a curve of a function w _i,j in an embodiment of the present invention;

FIG. 7 is a schematic diagram of a curve of a function edge_gain _i,j according to an embodiment of the present invention; FIG.

FIG. 8 is a schematic structural diagram of a video denoising apparatus according to an embodiment of the present invention.

detailed description

The technical solutions in the embodiments of the present invention will be clearly and completely described in the following with reference to the accompanying drawings.

The embodiment of the present invention provides a video denoising method, which is applied to a video denoising device, and the device may be disposed in a smart phone, a tablet computer, a smart TV, a multimedia player, etc., which is not limited by the present invention.

Referring to Figure 1, the method includes:

S101: detecting an input image of the target video, and obtaining a mosquito noise probability of each pixel in the input image;

In the implementation process, as shown in FIG. 2, S101 may include:

S201: detecting an input image of the target video, and obtaining a pixel type of each pixel in the input image;

In practical applications, S201 includes: performing gradient detection on the input image to obtain a gradient value of each pixel of the input image; performing local edge detection on the input image to obtain an edge information value of each pixel of the input image; and based on the gradient value and The edge information value determines the pixel type of each pixel of the input image.

Here, the gradient detection process is mainly used to detect the gradient of the current point, and the local edge detection process is mainly used to detect whether there are pixels around the current pixel that are different from the current pixel. The above two processes may be performed simultaneously or sequentially, and the invention is not limited.

First, introduce the gradient detection process.

For example, the gradient detection template is a 3×3 template as shown in FIG. 3, and the gradient detection output is recorded as grad, and each pixel corresponds to a gradient value. Then, for the i-th pixel in the above template, that is, the gradient value grad _0,0 of the pixel value x _0,0 of the center point (0,0) can be obtained by the following formula (1):

Where MAX represents the maximum value of all data, x _i,j represents the pixel value of the peripheral pixel of the (i,j) offset relative to the center point (0,0) in the template, abs(x _0,0 -x _{i, j} ) is the absolute value of the offset from the relative center point (0, 0) in the template.

It can be understood that the 3×3 template is only an example, and any other size template can be used as a template, which is not limited by the present invention.

Next, the local edge detection process is introduced.

For example, the local edge detection template is still an 11×11 template as shown in FIG. 4, and the local edge detection output is recorded as edge, and each pixel corresponds to an edge information value, then, for the ith pixel in the template, The edge information value edge _0,0 of the pixel value x _0,0 of the center point (0,0) can be obtained by the following formula (2).

Where MAX represents the maximum value of all data, x _i,j represents the pixel value of the peripheral pixel of the (i,j) offset relative to the center point (0,0) in the template, abs(x _0,0 -x _{i, j} ) is the absolute value of the offset from the relative center point (0, 0) in the template.

It should be noted that any template of other sizes can be used as a template when performing local edge detection, and preferably larger than the size of the template during gradient detection.

Then, after obtaining the gradient value and the edge information value of each pixel, based on these two values To determine the pixel type of each pixel.

Here, the pixel types are classified into three types, that is, a flat pixel (FLAT), a detail pixel (TEXTURE), and an edge pixel (EDGE). Then, when the gradient value of the ith pixel obtained by the above process is greater than or equal to the i-th pixel When the product of the edge information value and the first preset value is determined, the pixel type of the ith pixel is determined as an edge pixel; when the gradient value of the ith pixel is smaller than the edge information value of the ith pixel and the first preset value When the product is greater than or equal to the product of the edge information value of the ith pixel and the second preset value, the pixel type of the ith pixel is determined as the detail pixel; when the gradient value of the ith pixel is smaller than the ith pixel When the product of the edge information value and the first preset value is smaller than the product of the edge information value of the ith pixel and the second preset value, the pixel type of the ith pixel is determined as a flat pixel.

Here, the first preset value is a lower limit of a gradient value of the pixel and an edge information value ratio when the pixel type is an edge pixel, and the second preset value is a gradient value and an edge information value of the pixel when the pixel type is a detail pixel. The lower limit of the ratio. The first preset value is different from the second preset value, and both of the parameters are system input parameters, which can be configured according to different requirements of the user. Preferably, the first preset value is configured to be 0.6, and the second preset value is configured. Configured to be 0.2.

For example, the pixel type of each pixel is denoted as type, and then the pixel type type _m,n of the pixel whose image space coordinate is (m, n) can be obtained by the formula (3).

Where reg_edge_ratio represents the first preset value, and reg_text_ratio represents the second preset value. grad _{i, j} represents the relative center point (0,0) is offset within the template (i, j) of gradient values of peripheral _pixels, edge _{i, j} represents the template relative to a center point (0,0) is offset The edge information value of the peripheral pixels of (i, j).

S202: performing mosquito on each pixel in the input image based on the pixel type of each pixel The noise probability estimate is obtained to obtain the mosquito noise probability per pixel.

In an embodiment of the present invention, after the pixel type of each pixel obtained by S201, when the mosquito noise probability is performed on the current pixel, first, M pixels in a certain range around the current pixel are taken, and then, The M+1 pixels, that is, the pixel types of the M pixels around the i-th pixel and the ith pixel are counted, and then, based on the statistical result, the mosquito noise probability of the i-th pixel is determined. If the obtained mosquito noise probability value is larger, it means that the i-th pixel is more likely to be mosquito noise, and conversely, the possibility that the i-th pixel is mosquito noise is smaller.

For example, the template used for mosquito noise probability estimation is a 5×5 template as shown in FIG. 5, and the mosquito noise probability output is recorded as probability, and each pixel corresponds to a mosquito noise probability. Then, the mosquito noise probability value posibility _0,0 of the pixel value x _0,0 of the i-th pixel in the above template, that is, the center point (0, 0) can be obtained by the following formula (4).

Where type _{i, j} represents the pixel type of the pixel whose relative center point (0, 0) offset is (i, j) in the template, and type _{i, j} == FLAT represents the relative center point (0, 0) in the template. The pixel type of the pixel whose offset is (i, j) is a flat pixel, and type _{i, j} == TEXTURE indicates the pixel type of the pixel whose offset is (i, j) relative to the center point (0, 0) in the template. For the detail pixel, type _{i, j} == EDGE indicates that the pixel type of the pixel whose relative center point (0, 0) offset is (i, j) within the template is the edge pixel.

It can be understood that the 5×5 template is only an example, and any other size template can be used as a template, which is not limited by the present invention.

S102: Perform low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel.

In other embodiments of the present invention, S102 may also be executed in parallel while S101 is being executed. The input image is subjected to bilateral low-pass filtering to obtain a filtered image, so that the filtered pixel value corresponding to each pixel is obtained.

In other embodiments of the present invention, in order to preserve the detail edge of the input image and remove the mosquito noise, the S102 may further include: performing low-pass bilateral filtering based on the edge information value of each pixel and the pixel value of the input image. And obtain the filtered pixel value corresponding to each pixel.

For example, the low-pass bilateral filtering template is a 3×3 template as shown in FIG. 3, and the low-pass bilateral filtering output is recorded as flt, and each pixel corresponds to a filtered pixel value. Then, the filtered pixel value flt _0,0 of the pixel value x _0,0 of the i-th pixel in the above template, that is, the center point (0, 0) can be obtained by the following formula (5).

Where x _i,j represents the domain pixel with the offset of the relative center point (0,0) in the template as (i,j), and w _i,j represents the difference between the domain pixel and the center point (0,0) in the template. The function of the absolute value of the value, edge_gain _i,j is the edge intensity gain of the domain pixel with the offset of (i,j) relative to the center point (0,0) in the above template, which is obtained during the local edge detection process.

In the implementation, w _i,j can be obtained by the following formula (6), and can also refer to the curve as shown in FIG. 6.

是将

的值限制在区间[0,1]内。上述reg_diff_low和reg_diff_high为系统输入参数，可根据用户的需求进行配置，较优地，reg_diff_low配置为30，reg_diff_high 配置为100。Where reg_diff_low represents the lower limit of the absolute value of the difference between the two neighborhood pixels, and reg_diff_high represents the upper limit of the absolute value of the difference between the two neighborhood pixels.

Is going to

The value is limited to the interval [0,1]. The above reg_diff_low and reg_diff_high are system input parameters, which can be configured according to user requirements. Preferably, reg_diff_low is configured as 30 and reg_diff_high is configured as 100.

Further, the above function CLIP(h, low, high) can be implemented by the following formula (7).

Further, since the 3×3 template as shown in FIG. 3 is used, the edge_gain _{i, j} of all the neighborhood pixels in the template are equal, which is equal to the edge intensity gain of the center point (0, 0). Then, edge_gain _i,j can be obtained by the following formula (8), and can also refer to the curve shown in FIG.

Where reg_gain_thr_low represents the lower limit of the edge information value of the neighborhood pixel, reg_gain_thr_high represents the upper limit of the edge information value of the neighborhood pixel, and edge ₀ , ₀ represents the edge intensity gain of the center point (0, 0) in the template.

The above reg_gain_thr_low and reg_gain_thr_high are system input parameters, which can be configured according to user requirements. Preferably, reg_gain_thr_low is configured as 128, and reg_gain_thr_high is configured as 256.

In an actual application, the algorithm for low-pass filtering the input image may be other low-pass bilateral filtering algorithms in addition to the above-mentioned low-pass bilateral filtering algorithm, which is not limited by the present invention.

In summary, the filtered pixel values corresponding to each pixel can be obtained through the above steps.

S103: Perform weighting processing on the filtered pixel value and the pixel value of the input image based on the mosquito noise probability, determine a pixel value of the output image, and output an image;

In other embodiments of the present invention, the filtered pixel values and the pixel values of the input image are weighted by the following formula (9).

Xout _m,n =probability _m,n ×flt _m,n +(1-probability _m,n )×x _m,n (9);

Where xout _m,n represents the pixel value of the output image with the image space coordinate of (m,n), probability _m,n represents the mosquito noise probability value of the image space coordinate of (m,n), x _m,n represents the image The pixel value of the input image whose spatial coordinate is (m, n), flt _{m, n} represents the filtered pixel value of the image space coordinate of (m, n).

Then, after the weighting process described above, the pixel value of the output image corresponding to each pixel is obtained, and finally, the output image is output.

At this point, the denoising process of one frame of the input image in the target video is completed. For each frame of the video, the above steps are performed to perform denoising, which will not be repeated here.

It can be seen from the above that for each frame of the target video, only the pixels with relatively high mosquito noise probability are filtered, and other pixels are not processed, so that not only the mosquito noise existing in the video can be removed, but also It also preserves the details in the video to enhance the viewer's visual experience and video quality.

Based on the same inventive concept, an embodiment of the present invention further provides a video denoising apparatus, which is consistent with the video denoising apparatus described in one or more of the above embodiments.

As shown in FIG. 8, the apparatus includes: a detecting unit 1, a filtering unit 2, and an output unit 3; wherein the detecting unit 1 is configured to detect an input image of the target video to obtain a mosquito noise probability of each pixel in the input image; The filtering unit 2 is configured to perform low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; and the output unit 3 is configured to filter the pixel value and the pixel of the input image based on the mosquito noise probability The value is weighted to obtain a pixel value of the output image, and the output image is output.

In an embodiment of the present invention, the detecting unit 1 includes: a pixel type detecting unit configured to detect an input image of the target video, obtain a pixel type of each pixel in the input image; and a noise probability estimating unit configured to be based on each A pixel type of a pixel, which performs a mosquito noise probability estimation for each pixel in the input image to obtain a mosquito noise probability per pixel.

In an embodiment of the present invention, the pixel type detecting unit includes: a gradient detecting unit configured to perform gradient detection on the input image to obtain a gradient value of each pixel of the input image; and an edge detecting unit configured to input the image Performing local edge detection to obtain an edge information value of each pixel of the input image; and a pixel type determining unit configured to determine a pixel type of each pixel of the input image based on the gradient value and the edge information value.

In an embodiment of the present invention, the pixel type determining unit is configured to: when the gradient value of the ith pixel is greater than or equal to the product of the edge information value of the ith pixel and the first preset value, the ith pixel The pixel type is determined as an edge pixel, where i is a positive integer; and is configured to when the gradient value of the ith pixel is smaller than the product of the edge information value of the ith pixel and the first preset value, and is greater than or equal to the ith When the product of the edge information value of the pixel and the second preset value, the pixel type of the ith pixel is determined as the detail pixel, wherein the first preset value is different from the second preset value; and is configured to be the ith When the gradient value of the pixel is smaller than the product of the edge information value of the ith pixel and the first preset value, and is smaller than the product of the edge information value of the ith pixel and the second preset value, the pixel type of the ith pixel Determined to be a flat pixel.

In an embodiment of the present invention, the filtering unit 2 is configured to perform low-pass bilateral filtering based on the edge information value of each pixel and the pixel value of the input image to obtain a filtered pixel value corresponding to each pixel.

In an embodiment of the present invention, the noise probability estimating unit includes: a pixel type statistic unit configured to perform statistics on pixel types of the ith pixel and the M neighboring pixels, wherein the neighboring pixels are the ith a pixel around the pixel; a probability calculation unit configured to determine a mosquito noise probability of the ith pixel based on the statistical result.

In an embodiment of the present invention, the probability calculation unit is configured to calculate, according to the ith pixel and the M pixels, the number of pixels whose pixel type is a flat pixel, the number of pixels whose pixel type is a detail pixel, and the pixel type is an edge pixel. The ratio of the sum of the number of pixels determines the mosquito noise probability of the ith pixel.

It should be noted that each unit included in the device can be implemented by a processor in the terminal, and can also be implemented by a logic circuit; in the process of implementation, the processor can be a central processing unit (CPU). , microprocessor (MPU), digital signal processor (DSP) or field programmable gate array (FPGA). In addition, the description of the above device embodiments is similar to the description of the foregoing method embodiments, and has similar advantageous effects as the method embodiments, and thus will not be described again. For the technical details not disclosed in the embodiment of the present invention, please refer to the description of the method embodiment of the present invention, and the details are not described herein.

It should be noted that, in the embodiment of the present invention, if the video denoising method described above is implemented in the form of a software function module and sold or used as a standalone product, it may also be stored in a computer readable storage medium. Based on such understanding, the technical solution of the embodiments of the present invention may be embodied in the form of a software product in essence or in the form of a software product stored in a storage medium, including a plurality of instructions. A computer device (which may be a personal computer, server, or network device, etc.) is caused to perform all or part of the methods described in various embodiments of the present invention. The foregoing storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read only memory (ROM), a magnetic disk, or an optical disk. Thus, embodiments of the invention are not limited to any specific combination of hardware and software.

Correspondingly, the embodiment of the present invention further provides a computer storage medium, where the computer storage medium stores computer executable instructions, and the computer executable instructions are used to perform a video denoising method in the embodiment of the present invention.

Correspondingly, the embodiment of the present invention further provides a terminal (such as a computer, a smart phone, a tablet), and the terminal includes:

a storage medium configured to store computer executable instructions;

a processor configured to detect an input image of the target video to obtain a mosquito noise probability of each pixel in the input image;

Performing low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel;

And weighting the filtered pixel value and the pixel value of the input image according to the mosquito noise probability, obtaining a pixel value of the output image, and outputting the output image.

Those skilled in the art will appreciate that embodiments of the present invention can be provided as a method, system, or computer program product. Accordingly, the present invention can take the form of a hardware embodiment, a software embodiment, or a combination of software and hardware. Moreover, the invention can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) including computer usable program code.

The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (system), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or FIG. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine for the execution of instructions for execution by a processor of a computer or other programmable data processing device. Means for implementing the functions specified in one or more of the flow or in a block or blocks of the flow chart.

The computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture comprising the instruction device. The apparatus implements the functions specified in one or more blocks of a flow or a flow and/or block diagram of the flowchart.

These computer program instructions can also be loaded onto a computer or other programmable data processing device such that a series of operational steps are performed on a computer or other programmable device to produce computer-implemented processing for execution on a computer or other programmable device. The instructions are provided to implement the work specified in one or more blocks of a flow or a flow and/or a block diagram of the flowchart The steps that can be made.

The above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention.

Industrial applicability

In the embodiment of the present invention, the input image of the target video is detected, the mosquito noise probability of each pixel in the input image is obtained, and the input image is low-pass filtered to obtain the filtered pixel value corresponding to each pixel; And performing weighting processing on the filtered pixel value and the pixel value of the input image based on the mosquito noise probability, obtaining a pixel value of the output image, and outputting the output image. That is to say, for each frame of the target video, only pixels with a relatively high mosquito noise probability are filtered, and other pixels are not processed, so that not only the mosquito noise existing in the video but also the mosquito noise can be removed. The details in the video are preserved to enhance the viewer's visual experience and video quality.

Claims (16)

A video denoising method includes: Detecting an input image of the target video to obtain a mosquito noise probability of each pixel in the input image; Performing low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; And weighting the filtered pixel value and the pixel value of the input image according to the mosquito noise probability, obtaining a pixel value of the output image, and outputting the output image. The method of claim 1, wherein the detecting an input image of the target video to obtain a mosquito noise probability of each pixel in the input image comprises: Detecting an input image of the target video to obtain a pixel type of each pixel in the input image; A mosquito noise probability estimation is performed on each of the pixels in the input image based on a pixel type of each of the pixels, and a mosquito noise probability of each of the pixels is obtained. The method of claim 2, wherein the detecting an input image of the target video to obtain a pixel type of each pixel in the input image comprises: Performing gradient detection on the input image to obtain a gradient value of each pixel of the input image; Performing local edge detection on the input image to obtain edge information values of each pixel of the input image; A pixel type of each pixel of the input image is determined based on the gradient value and the edge information value. The method of claim 3, wherein the determining a pixel type of each pixel of the input image based on the gradient value and the edge information value comprises: When the gradient value of the ith pixel is greater than or equal to the edge information of the ith pixel When the value is a product of the first preset value, the pixel type of the ith pixel is determined as an edge pixel, where i is a positive integer; When the gradient value of the ith pixel is smaller than the product of the edge information value of the ith pixel and the first preset value, and is greater than or equal to the edge information value of the ith pixel Determining, by the product of the second preset value, the pixel type of the ith pixel as a detail pixel, wherein the first preset value is different from the second preset value; When the gradient value of the ith pixel is smaller than a product of the edge information value of the ith pixel and the first preset value, and smaller than the edge information value of the ith pixel When the product of the second preset value is combined, the pixel type of the ith pixel is determined to be a flat pixel. The method of claim 3, wherein the low-pass filtering the input image to obtain the filtered pixel value corresponding to each pixel comprises: And performing low-pass bilateral filtering on the edge information value of each of the pixels and the pixel value of the input image to obtain a filtered pixel value corresponding to each pixel. The method according to claim 2, wherein said mosquito noise probability estimation is performed on said each pixel in said input image based on a pixel type of said each pixel, and said mosquito noise of said each pixel is obtained Probability, including: Counting pixel types of the ith pixel and the M neighboring pixels, wherein the neighboring pixels are pixels around the ith pixel; Based on the statistical result, the mosquito noise probability of the ith pixel is determined. The method according to claim 6, wherein the determining a mosquito noise probability of the ith pixel based on a statistical result comprises: Determining the i-th based on a ratio of a number of pixels in which the pixel type is a flat pixel in the ith pixel and the M pixels, and a ratio of a pixel number of the pixel type to the detail pixel and a pixel type of the edge type The probability of mosquito noise in pixels. A video denoising device includes: a detecting unit, a filtering unit, and an output unit; among them, The detecting unit is configured to detect an input image of the target video, and obtain a mosquito noise probability of each pixel in the input image; The filtering unit is configured to perform low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; The output unit is configured to perform weighting processing on the filtered pixel value and the pixel value of the input image based on the mosquito noise probability, obtain a pixel value of the output image, and output the output image. The apparatus according to claim 8, wherein the detecting unit comprises: a pixel type detecting unit configured to detect an input image of the target video to obtain a pixel type of each pixel in the input image; The noise probability estimating unit is configured to perform a mosquito noise probability estimation on each of the pixels in the input image based on a pixel type of each of the pixels, to obtain a mosquito noise probability of each of the pixels. The device of claim 9, wherein the pixel type detecting unit comprises: a gradient detecting unit configured to perform gradient detection on the input image to obtain a gradient value of each pixel of the input image; An edge detecting unit configured to perform local edge detection on the input image to obtain an edge information value of each pixel of the input image; A pixel type determining unit is configured to determine a pixel type of each pixel of the input image based on the gradient value and the edge information value. The apparatus according to claim 10, wherein the pixel type determining unit is configured to: when the gradient value of the ith pixel is greater than or equal to the edge information value of the ith pixel and a first preset value When the product is the same, the pixel type of the ith pixel is determined as the edge image Is a positive integer; configured to: when the gradient value of the ith pixel is smaller than a product of the edge information value of the ith pixel and a first preset value, and greater than or equal to Determining, by the product of the edge information value of the i-th pixel, a pixel type of the i-th pixel as a detail pixel, wherein the first preset value is different from the second a preset value; the method further configured to: when the gradient value of the ith pixel is smaller than a product of the edge information value of the ith pixel and the first preset value, and smaller than the ith When the product of the edge information value of the pixel and the second preset value is used, the pixel type of the ith pixel is determined to be a flat pixel. The apparatus according to claim 11, wherein the filtering unit is configured to perform low-pass bilateral filtering based on edge information values of each pixel and pixel values of the input image to obtain corresponding to each pixel The filtered pixel value. The apparatus according to claim 10, wherein said noise probability estimating unit comprises: a pixel type statistic unit configured to count pixel types of the ith pixel and the M neighboring pixels, wherein the neighborhood pixel is a pixel around the ith pixel; The probability calculation unit is configured to determine a mosquito noise probability of the ith pixel based on the statistical result. The apparatus according to claim 13, wherein the probability calculation unit is configured to determine, according to the ith pixel and the M pixels, a pixel type that is a flat pixel, and a number of pixels whose pixel type is a detail pixel The ratio of the mosquito noise of the ith pixel is determined in proportion to the sum of the number of pixels whose pixel type is an edge pixel. A computer storage medium having stored therein computer executable instructions for performing the video denoising method of any one of claims 1 to 7. A terminal, the terminal comprising: a storage medium configured to store computer executable instructions; a processor configured to detect an input image of the target video to obtain a mosquito noise probability of each pixel in the input image; Performing low-pass filtering on the input image to obtain a filtered pixel value corresponding to each pixel; And weighting the filtered pixel value and the pixel value of the input image according to the mosquito noise probability, obtaining a pixel value of the output image, and outputting the output image.

Images