WO2018099136A1

WO2018099136A1 - Method and device for denoising image with low illumination, and storage medium

Info

Publication number: WO2018099136A1
Application number: PCT/CN2017/097318
Authority: WO
Inventors: 刘浩
Original assignee: 深圳市中兴微电子技术有限公司
Priority date: 2016-11-29
Filing date: 2017-08-14
Publication date: 2018-06-07
Also published as: CN108122206A

Abstract

Disclosed in the present invention is a method for denoising an image with low illumination, comprising: obtaining an image with low illumination; according to the difference of image texture information, segmenting the image with low illumination into different texture regions and dividing the texture regions into a foreground image region and a background image region; respectively denoising the foreground image region and the background image region by using different processing strategies to obtain a processed foreground image region and background image region; and combining the processed foreground image region and background image region to obtain a denoised image. Further disclosed in the present invention is a device for denoising an image with low illumination at the same time.

Description

Low illumination image denoising method, device and storage medium

Cross-reference to related applications

The present application is based on a Chinese patent application filed on Jan. 29, 2016, the entire disclosure of which is hereby incorporated by reference.

Technical field

The present invention relates to image processing technologies, and in particular, to a low illumination image noise reduction method, apparatus, and storage medium.

Background technique

In recent years, with the popularization of networks and computers, the application of video surveillance equipment to urban public security and smart homes has shown a rapid development trend. At the same time, people have higher and higher requirements for image quality obtained by video surveillance, that is, they want to obtain high-definition, low-stream video images. However, low illumination, such as night, backlight, indoor and other conditions, will greatly reduce the performance of video surveillance equipment, so that the video visibility obtained by video surveillance equipment is reduced, making it difficult to distinguish key people or things in the image. Generally, an image taken in the above low-light scene is referred to as a low-illumination image, and since the low-illumination image contains a large amount of noise, not only the image quality is lowered, but also the recognition of the image by the human eye is seriously affected, and the intelligent traffic and the target are also caused. The performance of computer vision systems that recognize and process based on low illumination images is greatly affected.

In order to improve the visual effect of the image, the image enhancement process is usually performed on the low-illumination image, highlighting the interesting part of the image, enhancing the useful information in the image, and weakening or removing the unnecessary information, so that the useful information is strengthened, thereby obtaining A more practical image or converted into an image that is more suitable for human or machine analysis processing. However, for traditional image enhancement techniques than If the image is processed by the technique, the original noise in the image will be greatly enhanced, and a large amount of color noise and some brightness noise often appear in the image. Therefore, the noise reduction processing of low illumination images is an urgent problem to be solved.

For the traditional noise reduction method, the separate spatial filtering can filter out some noise, but it is easy to cause loss of image detail or block effect, especially in the H264 video coding with large quantization parameter (QP, Quantization Parameter). The image encoded by spatial filtering is prone to become ambiguous; although the time domain filtering alone can make good use of the correlation between video frames, the noise existing in the still image can be filtered out, but for the motion in the image. Objects can cause serious "tailing" phenomenon; compared with spatial filtering and time domain filtering alone, space-time domain joint filtering can suppress noise to some extent, but after filtering the pixel space, it will make the image The edge portion becomes less noticeable, causing the display of the entire image to become blurred so that the details of the edge of the image are lost. It can be seen that the above existing noise reduction methods can not guarantee the color saturation of the object while filtering out the image noise, and it is difficult to preserve the details of the object while filtering the noise in the low illumination scene.

Summary of the invention

In view of the above, the embodiments of the present invention are directed to providing a low-illuminance image denoising method, apparatus, and storage medium, which are intended to solve the above problems existing in the low-illuminance image processing by the existing noise reduction method, and can effectively reduce image noise. Maximize the original information of the image.

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

Embodiments of the present invention provide a low illumination image denoising method, where the method includes:

Obtaining a low illumination image;

Dividing the low-illumination image into different texture regions according to different image texture information, and dividing the texture region into a foreground image region and a background image region;

Performing noise reduction processing on the foreground image area and the background image area respectively to obtain a processed foreground image area and a background image area;

The processed foreground image area and the background image area are combined to obtain a noise reduction image.

In the above solution, the foreground image area and the background image area each include a component signal; the component signal includes: a luminance component signal and a chrominance component signal.

In the above solution, before the obtaining the low illumination image, the method further comprises: training the reference background image for learning the normal illumination intensity.

In the above solution, after the dividing the low illumination image into different texture regions, the method further comprises: calculating a standard deviation of pixel gray levels in each texture region;

Performing noise reduction processing on the foreground image area, including:

Determining, according to a standard deviation of pixel gradations in each texture region, a filter coefficient corresponding to each texture region of the foreground image region; respectively filtering the luminance component signals of each texture region by using respective corresponding filter coefficients to obtain a first a noise reduction component image; extracting a texture structure according to different texture partitions of the foreground image to obtain a texture sharpening image; superimposing the first noise reduction component image and the texture sharpening image to obtain a second noise reduction component image ;

According to different texture regions, the chrominance component signals of the foreground image region are subjected to sub-region filtering by using different filter coefficients.

In the above solution, the performing noise reduction processing on the background image area includes:

Performing time domain filtering on the luminance component signal of the background image region;

A chrominance component signal of the background image region is temporally filtered based on the reference background image of the normal illumination intensity.

The embodiment of the present invention further provides a low-intensity image noise reduction device, where the device includes: an image acquisition module area division module and a noise reduction processing module;

The image acquisition module is configured to acquire a low illumination image;

The area dividing module is configured to divide the low-illumination image into different texture regions according to different image texture information, and divide the texture region into a foreground image region and a back Scene image area;

The noise reduction processing module is configured to perform noise reduction processing on the foreground image area and the background image area respectively to obtain a processed foreground image area and a background image area, and after the processing The foreground image area and the background image area are combined to obtain a noise reduction image.

In the above solution, the apparatus further includes: a training learning module configured to train the reference background image for learning the normal illumination intensity before the image acquisition module acquires the low illumination image.

In the above solution, the device further includes: a calculating module, configured to calculate, after the region dividing module divides the low-illumination image into different texture regions, calculate a standard deviation of pixel gray levels in each texture region;

The noise reduction processing module includes: a foreground luminance component signal processing module and a foreground color component signal processing module; wherein

The foreground luminance component signal processing module is configured to determine, according to a standard deviation of pixel gray levels in each texture region, a filter coefficient corresponding to each texture region of the foreground image region; respectively adopting respective corresponding filter coefficients for each texture The luminance component signal of the region is filtered to obtain a first noise reduction component image; the texture structure is extracted according to different texture partitions of the foreground image to obtain a texture sharpening image; and the first noise reduction component image and the texture are sharpened The images are superimposed to obtain a second noise reduction component image;

The foreground color component signal processing module is configured to perform sub-region filtering on the chrominance component signals of the foreground image region by using different filter coefficients according to different texture regions.

In the above solution, the noise reduction processing module further includes: a background luminance component signal processing module, and a background chrominance component signal processing module; wherein

The background luminance component signal processing module is configured to brightness the background image region Component signal is time domain filtered;

The background chrominance component signal processing module is configured to perform temporal filtering on the chrominance component signal of the background image region based on the reference background image of the normal illumination intensity.

The embodiment of the invention further provides a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is implemented by the processor to implement the low illumination image denoising method.

Embodiments of the present invention also provide a low illumination image noise reduction device, a processor, and a memory for storing a computer program capable of running on a processor,

Wherein the processor is configured to perform the low illumination image denoising method when the computer program is executed.

The low illumination image denoising method, device and storage medium provided by the embodiments of the present invention acquire a low illumination image; according to different image texture information, the low illumination image is segmented into different texture regions, and the texture region is Dividing into a foreground image area and a background image area; performing different noise reduction processing on the foreground image area and the background image area respectively, obtaining a processed foreground image area and a background image area, and performing the processing The subsequent foreground image area and the background image area are combined to obtain a noise reduction image. In this way, taking full advantage of the characteristics of low-light images, combining noise reduction technology with texture sharpening technology can effectively reduce image noise, and in the process of filtering out noise, try to ensure the color saturation of the object and reduce the details of the object. Loss to maximize the original information of the image and improve the user's subjective feelings.

DRAWINGS

1 is a schematic flowchart of a low illumination image denoising method according to Embodiment 1 of the present invention;

2 is a schematic flowchart of a specific implementation process of a low illumination image denoising method according to Embodiment 2 of the present invention;

3 is a schematic structural diagram of a low illumination image noise reduction device according to Embodiment 3 of the present invention;

FIG. 4 is a schematic structural diagram of hardware components of a low illumination image noise reduction device according to an embodiment of the present invention; Figure.

detailed description

The embodiments of the present invention are described in detail with reference to the accompanying drawings.

Embodiment 1

As shown in FIG. 1 , the implementation process of the low illumination image denoising method in the embodiment of the present invention includes the following steps:

Step 101: Acquire a low illumination image;

Before performing this step, the method further includes: training a reference background image that learns normal light intensity.

Here, first, according to the image motion estimation algorithm, the moving object and the stationary object in the image under normal illumination intensity are separated. Generally, the area where the moving object is located is referred to as the foreground image area, and the area where the stationary object is located is called As the background image area; then, the reference background image learning normal light intensity is trained using an intelligent learning algorithm. Preferably, the intelligent learning algorithm may adopt a k-means clustering algorithm in a neural network learning algorithm; wherein, in the case of good light, the reference background image will automatically perform background update.

Here, in image or video processing applications, motion estimation is a very basic pre-processing step, but the amount of motion estimation is very large. Therefore, it is important to perform parallel processing on the motion estimation algorithm. A motion estimation algorithm that separates moving objects in an image. For example, how to separate the moving object and the stationary object in the image under normal illumination intensity according to the image motion estimation algorithm belongs to the prior art, and will not be further described herein.

Here, it should be noted that the surveillance camera position needs to be fixed in advance, and the image under normal light intensity and the low illumination image are acquired at the same shooting angle at the same position.

Step 102: Segment the low illumination image into different texture regions according to different image texture information, and divide the texture region into a foreground image region and a background image region;

Here, the existing super pixel segmentation technique may be adopted, and the low illumination image is segmented into different texture regions according to different image texture information; the texture region may be divided into foreground image regions and according to an image motion estimation algorithm. Background image area;

After the step of dividing the low-illumination image into different texture regions, the method further includes: calculating a standard deviation of pixel gray levels in each texture region; wherein a standard deviation of the pixel gray levels belongs to One of the feature information of the texture area;

The foreground image area and the background image area each include a component signal; the component signal includes: a luminance component signal and a chrominance component signal.

Further, after the low-illumination image is segmented into different texture regions, the method further includes: extracting feature information of each texture region; wherein the feature information of each texture region includes: pixel points in each texture region The standard deviation of the gray scale, and the gradient value of the gray level of the pixel in each texture region.

Here, the standard deviation of the pixel point gradation in each texture region is taken as the texture and noise level parameter of the image, so that the image is weighted and averaged according to the texture and noise level parameters of the image.

Step 103: Perform different noise reduction processing on the foreground image area and the background image area respectively to obtain a processed foreground image area and a background image area.

Here, the performing noise reduction processing on the foreground image area includes:

Here, the filter coefficient corresponding to each texture region of the foreground image region may be determined according to a mapping relationship between a standard deviation of pixel gradation and a filter coefficient in each texture region; wherein the determining process belongs to the prior art, where I will not repeat them one by one.

In a preferred embodiment, different filter coefficients are respectively used, and the luminance component signals of the texture regions are filtered by using a three-dimensional block matched filter (BM3D) to obtain a first noise reduction component image; according to different textures of the foreground image The partitioning can further determine the brightness and darkness of each texture region, and then select a suitable filter coefficient according to the brightness of each texture region, and extract the texture structure of the foreground image. The main texture image of the noise, that is, the texture sharpening image; wherein, for the texture region with higher brightness, a smaller filter coefficient is selected, and for the texture region with lower brightness, a relatively larger filter coefficient is selected, that is, The filter coefficient may be an empirical value selected by the user according to experience. Finally, the first noise reduction component image is superimposed with the texture sharpening image to obtain a second noise reduction component image to complete the noise reduction processing on the foreground image luminance component signal. For the foreground image chrominance component signal, the chrominance component signals of the foreground image region are subjected to sub-region filtering according to the brightness degree of each texture region of the image, so as to achieve color restoration of the low illuminance image.

Here, the purpose of sharpening is to highlight the edge contour of the object and facilitate object recognition. Common algorithms include gradient method, operator, high-pass filtering, mask matching method, and statistical difference method.

It is important to note that filtering with BM3D preserves the texture details of the image as much as possible while removing image noise.

Here, the performing noise reduction processing on the background image area includes:

In a preferred embodiment, since the scene scene in the background image area is relatively fixed, the effect of noise reduction by using time domain filtering is better. That is, it is usually to continuously take several images of the same scene, and then perform weighted averaging processing on the images according to the texture and noise level parameters of the image, so that an image with relatively relatively low noise can be obtained.

Step 104: Synthesize the processed foreground image region and the background image region to obtain a noise reduction image.

Here, the processed foreground image region and the background image region may be synthesized by a program using existing weighted overlay synthesis, semi-transparent overlay synthesis or gradient synthesis to obtain a complete effect map.

In the embodiment of the present invention, the low-illumination video surveillance image is divided into a background area and a foreground area according to different image texture information, and different processing strategies are adopted for the background area and the foreground area to perform noise reduction processing, and after the noise reduction processing is completed The background area and the foreground area are combined, and the synthesized image is taken as the finally obtained noise reduction image. In this way, image noise can be effectively reduced, and the original information of the image can be maximized. At the same time, the video image after noise reduction can not only reduce the code stream, but also make the code stream smooth, which is conducive to network transmission, and can also amplify the gain in low illumination scenes; at the same time, it is convenient for users to extract image feature information for subsequent video intelligence. Prepare for analysis.

Embodiment 2

The specific implementation process of the low illumination image denoising method in the embodiment of the present invention is further described in detail below.

FIG. 2 is a schematic diagram showing a specific implementation process of a low illumination image denoising method according to an embodiment of the present invention. As shown in FIG. 2, the method includes the following steps:

Step 201: Collect a monitoring image.

Step 205: According to the light sense, it is determined whether the current illumination environment belongs to normal illumination or a scene with low illumination, if it is normal illumination, step 203 is performed; if it is a low illumination scene, step 205 is performed;

Here, a light sensor can be used to detect light.

Step 203: Perform motion learning background learning on the image of normal illumination;

Here, the moving object and the stationary object in the image under normal illumination intensity may be separated according to the image motion estimation algorithm; generally, the region in which the moving object is located is referred to as a foreground image region, and the region in which the stationary object is located is referred to as Background image area.

Here, any one of the neural network learning algorithms may be employed to perform training learning of the motion detection background on the normally illuminated image to obtain a reference background image of normal illumination intensity. Preferably, the reference background image can be trained and learned using a k-means clustering algorithm.

Step 204: Obtain a reference background image of normal light intensity, and proceed to step 206;

Here, in the case where the light is good, the reference background image will always automatically update the background.

Step 205: Split the low illumination image into different texture regions;

Here, an existing super pixel segmentation technique may be employed, and the low illumination image is segmented into different texture regions according to different image texture information, and a standard deviation of pixel gray levels in each texture region is extracted; wherein The standard deviation of the gray level of the pixel belongs to one of the feature information of the texture region, and the feature information of the texture region further includes: a gradient value of the gray level of the pixel. Here, the standard deviation of the pixel point gradation in each texture region is taken as the texture and noise level parameter of the image, so that the image is weighted and averaged according to the texture and noise level parameters of the image.

Step 206: Divide the texture area into a foreground image area and a background image area based on the reference background image;

Here, the texture region may be divided into a foreground image region and a background image region according to an image motion estimation algorithm; the foreground image region and the background image region each include a component signal; the component signal includes: a luminance component signal and Chroma component signal.

Step 207: Perform noise reduction processing on the foreground image area and the background image area respectively.

Here, the BM3D is used to filter the luminance component signals of the texture regions, so that the texture details of the images can be preserved as much as possible while removing the image noise.

The time domain filtering generally takes several images continuously for the same scene, and then performs weighted averaging processing on the images according to the texture and noise level parameters of the image, so that an image with relatively relatively low noise can be obtained.

Step 208: Synthesize the foreground image area and the background image area after the noise reduction processing;

Step 209: Output a noise reduction video image.

In the embodiment of the present invention, the low-illumination video surveillance image is divided according to different image texture information. Cut into the background area and the foreground area, and adopt different processing strategies for the background area and the foreground area to perform noise reduction processing, and synthesize the background area and the foreground area after the noise reduction processing is completed, and the synthesized image is finally obtained. Noise reduction image. In this way, image noise can be effectively reduced, and the original information of the image can be maximized. At the same time, the video image after noise reduction can not only reduce the code stream, but also make the code stream smooth, which is conducive to network transmission, and can also amplify the gain in low illumination scenes; at the same time, it is convenient for users to extract image feature information for subsequent video intelligence. Prepare for analysis.

Embodiment 3

In order to achieve the above method, the embodiment of the present invention further provides a low-intensity image noise reduction device. As shown in FIG. 3, the device includes an image acquisition module 301, a region division module 302, and a noise reduction processing module 303.

The image obtaining module 301 is configured to acquire a low illumination image;

The area dividing module 302 is configured to divide the low-illumination image into different texture regions according to different image texture information, and divide the texture region into a foreground image region and a background image region;

The noise reduction processing module 303 is configured to perform noise reduction processing on the foreground image region and the background image region by using different processing strategies to obtain the processed foreground image region and the background image region, and the processing is performed. The subsequent foreground image area and the background image area are combined to obtain a noise reduction image.

Here, the apparatus further includes a training learning module 304 configured to train a reference background image that learns normal light intensity before the image acquisition module 301 acquires a low illumination image.

The apparatus further includes a calculation module 305 configured to calculate a standard deviation of pixel gradations in each texture region after the region dividing module 302 divides the low illuminance image into different texture regions.

Here, the noise reduction processing module 303 includes: a foreground luminance component signal processing module, and a foreground color component signal processing module;

The noise reduction processing module 303 further includes: a background luminance component signal processing module and a background chrominance component signal processing module; wherein

The background luminance component signal processing module is configured to perform time domain filtering on the luminance component signal of the background image region;

In an actual application, the image obtaining module 301, the area dividing module 302, the noise reduction processing module 303, the training learning module 304, and the computing module 305 may each be a central processing unit (CPU) located on the image processing terminal. Microprocessor (MPU), digital signal processor (DSP, Digital Signal Processor), or Field Programmable Gate Array (FPGA).

The embodiment of the present invention obtains a low illumination image; the low illumination image is segmented into different texture regions according to different image texture information, and the texture region is divided into a foreground image region and a background image region; and the foreground image is The area and the background image area are respectively subjected to noise reduction processing by using different processing strategies, and the processed foreground image area and the background image area are obtained. The processed foreground image region and the background image region are combined to obtain a noise reduction image. In this way, taking full advantage of the characteristics of low-light images, combining noise reduction technology with texture sharpening technology can effectively reduce image noise, and in the process of filtering out noise, try to ensure the color saturation of the object and reduce the details of the object. Loss to maximize the original information of the image and improve the user's subjective feelings.

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.

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.

Correspondingly, an embodiment of the present invention further provides a computer storage medium, where the computer storage medium stores a computer program, and when the computer program is executed by the processor, executes:

Obtaining a low illumination image;

The foreground image area and the background image area each include a component signal; the component signal includes a luminance component signal and a chrominance component signal.

When the computer program is executed by the processor, it also performs: training a reference background image that learns normal light intensity.

When the computer program is executed by the processor, it also executes:

Calculating a standard deviation of pixel gray levels in each texture region;

Performing noise reduction processing on the foreground image area, including:

When the computer program is executed by the processor, it also executes:

An embodiment of the present invention further provides a low illumination image noise reduction device, the composition of the low illumination image noise reduction device comprising: a processor and a memory for storing a computer program capable of running on a processor, wherein When the processor is used to run the computer program, execute:

Obtaining a low illumination image;

Different processing strategies are applied to the foreground image area and the background image area respectively Performing a noise reduction process to obtain a processed foreground image area and a background image area;

The processed foreground image area and the background image area are combined to obtain a noise reduction image. The foreground image area and the background image area each include a component signal; the component signal includes a luminance component signal and a chrominance component signal.

The processor is further configured to: when operating the computer program, perform: training a reference background image that learns normal light intensity.

The processor is further configured to execute when the computer program is executed:

Calculating a standard deviation of pixel gray levels in each texture region;

Performing noise reduction processing on the foreground image area, including:

4 is a schematic diagram showing the hardware composition of a low-light image noise reduction device according to an embodiment of the present invention. The server 700 includes at least one processor 701, a memory 702, and at least one network interface 704. The various components in server 700 are coupled together by a bus system 705. It will be appreciated that the bus system 705 is used to implement connection communication between these components. The bus system 705 includes a power bus, a control bus, and a status signal bus in addition to the data bus. But to be clear For clarity, various buses are labeled as bus system 705 in FIG.

It will be appreciated that memory 702 can be either volatile memory or non-volatile memory, and can include both volatile and nonvolatile memory. The non-volatile memory may be a ROM, a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), or an electrically erasable device. EEPROM (Electrically Erasable Programmable Read-Only Memory), Ferromagnetic random access memory (FRAM), Flash Memory, Magnetic Surface Memory, Optical Disk, or Read Only Disc (CD) -ROM, Compact Disc Read-Only Memory); the magnetic surface memory may be a disk storage or a tape storage. The volatile memory can be a random access memory (RAM) that acts as an external cache. By way of example and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access (SSRAM). DRAM (Dynamic Random Access Memory), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), enhancement Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Synchronous Dynamic Random Access Memory (SLDRAM), Direct Memory Bus Random Access Memory (DRRAM) ). The memory 702 described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

The memory 702 in the embodiment of the present invention is used to store various types of data to support the operation of the low illumination image noise reduction device 700. Examples of such data include: for noise reduction in low illumination images Any computer program, such as application 7022, operating on device 700. A program implementing the method of the embodiment of the present invention may be included in the application 7022.

The method disclosed in the foregoing embodiments of the present invention may be applied to the processor 701 or implemented by the processor 701. Processor 701 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the foregoing method may be completed by an integrated logic circuit of hardware in the processor 701 or an instruction in a form of software. The processor 701 described above may be a general purpose processor, a digital signal processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, or the like. The processor 701 can implement or perform the various methods, steps, and logic blocks disclosed in the embodiments of the present invention. A general purpose processor can be a microprocessor or any conventional processor or the like. The steps of the method disclosed in the embodiment of the present invention may be directly implemented as a hardware decoding processor, or may be performed by a combination of hardware and software modules in the decoding processor. The software module can reside in a storage medium located in memory 702, which reads the information in memory 702 and, in conjunction with its hardware, performs the steps of the foregoing method.

In an exemplary embodiment, the low-illumination image noise reduction device 700 may be one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), and complex programmable A logic device (CPLD, Complex Programmable Logic Device), FPGA, general purpose processor, controller, MCU, MPU, or other electronic component implementation for performing the aforementioned method.

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. Realize A means of function specified in a flow or a flow and/or a block diagram of a block or blocks.

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 provide steps for implementing the functions specified in one or more of the flow or in a block or blocks of a flow diagram.

The above is only the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in Within the scope of protection of the present invention.

Industrial applicability

In the embodiment of the present invention, a low illumination image is acquired; the low illumination image is segmented into different texture regions according to different image texture information, and the texture region is divided into a foreground image region and a background image region; The foreground image area and the background image area are respectively subjected to noise reduction processing by different processing strategies, and the processed foreground image area and the background image area are obtained, and the processed foreground image area and the background image area are combined to obtain Noise reduction image. In this way, taking full advantage of the characteristics of low-light images, combining noise reduction technology with texture sharpening technology can effectively reduce image noise, and in the process of filtering out noise, try to ensure the color saturation of the object and reduce the details of the object. Loss to maximize the original information of the image.

Claims

A low illumination image denoising method, the method comprising:

Obtaining a low illumination image;

Dividing the low-illumination image into different texture regions according to different image texture information, and dividing the texture region into a foreground image region and a background image region;

Performing noise reduction processing on the foreground image area and the background image area respectively to obtain a processed foreground image area and a background image area;

The processed foreground image area and the background image area are combined to obtain a noise reduction image.
The method of claim 1, wherein the foreground image area and the background image area each comprise a component signal; the component signal comprises: a luminance component signal and a chrominance component signal.
The method of claim 2, wherein before the obtaining the low illumination image, the method further comprises:

Train a reference background image that learns normal light intensity.
The method of claim 2, wherein after the dividing the low illumination image into different texture regions, the method further comprises:

Calculating a standard deviation of pixel gray levels in each texture region;

Performing noise reduction processing on the foreground image area, including:

Determining, according to a standard deviation of pixel gradations in each texture region, a filter coefficient corresponding to each texture region of the foreground image region; respectively filtering the luminance component signals of each texture region by using respective corresponding filter coefficients to obtain a first a noise reduction component image; extracting a texture structure according to different texture partitions of the foreground image to obtain a texture sharpening image; superimposing the first noise reduction component image and the texture sharpening image to obtain a second noise reduction component image ;

According to different texture regions, the chrominance component signal of the foreground image region is not used. The same filter coefficients are used for sub-region filtering.
The method according to claim 3, wherein said performing noise reduction processing on said background image region comprises:

Performing time domain filtering on the luminance component signal of the background image region;

A chrominance component signal of the background image region is temporally filtered based on the reference background image of the normal illumination intensity.
A low illumination image noise reduction device, the device comprising: an image acquisition module area division module and a noise reduction processing module; wherein

The image acquisition module is configured to acquire a low illumination image;

The area dividing module is configured to divide the low-illumination image into different texture regions according to different image texture information, and divide the texture region into a foreground image region and a background image region;

The noise reduction processing module is configured to perform noise reduction processing on the foreground image area and the background image area respectively to obtain a processed foreground image area and a background image area, and after the processing The foreground image area and the background image area are combined to obtain a noise reduction image.
The apparatus of claim 6, wherein the foreground image area and the background image area each comprise a component signal; the component signal comprises: a luminance component signal and a chrominance component signal.
The apparatus of claim 7, wherein the apparatus further comprises: a training learning module configured to train a reference background image that learns normal light intensity before the image acquisition module acquires a low illumination image.
The apparatus according to claim 7, wherein the apparatus further comprises: a calculation module configured to calculate a pixel point gray in each texture area after the area dividing module divides the low illumination image into different texture areas Standard deviation of degree;

The noise reduction processing module includes: a foreground luminance component signal processing module and a foreground color component signal processing module; wherein

The foreground luminance component signal processing module is configured to determine, according to a standard deviation of pixel gray levels in each texture region, a filter coefficient corresponding to each texture region of the foreground image region; respectively adopting respective corresponding filter coefficients for each texture The luminance component signal of the region is filtered to obtain a first noise reduction component image; the texture structure is extracted according to different texture partitions of the foreground image to obtain a texture sharpening image; and the first noise reduction component image and the texture are sharpened The images are superimposed to obtain a second noise reduction component image;

The foreground color component signal processing module is configured to perform sub-region filtering on the chrominance component signals of the foreground image region by using different filter coefficients according to different texture regions.
The apparatus according to claim 8, wherein the noise reduction processing module further comprises: a background luminance component signal processing module, and a background chrominance component signal processing module; wherein

The background luminance component signal processing module is configured to perform time domain filtering on the luminance component signal of the background image region;

The background chrominance component signal processing module is configured to perform temporal filtering on the chrominance component signal of the background image region based on the reference background image of the normal illumination intensity.
A computer storage medium having stored therein a computer program that, when executed by a processor, implements the steps of the method of claims 1 to 5.
A low illumination image noise reduction device, comprising: a processor and a memory for storing a computer program executable on the processor,

Wherein the processor is operative to perform the steps of the method of claims 1 to 5 when the computer program is run.