CN113850741A

CN113850741A - Image noise reduction method and device, electronic equipment and storage medium

Info

Publication number: CN113850741A
Application number: CN202111178701.0A
Authority: CN
Inventors: 严洪泽
Original assignee: Hangzhou Zhicun Intelligent Technology Co ltd
Current assignee: Hangzhou Zhicun Intelligent Technology Co ltd
Priority date: 2021-10-10
Filing date: 2021-10-10
Publication date: 2021-12-28
Anticipated expiration: 2041-10-10
Also published as: CN113850741B

Abstract

The embodiment of the invention provides an image noise reduction method, an image noise reduction device, electronic equipment and a storage medium, wherein the method comprises the following steps: processing the target RAW image by using a pre-acquired noise model to obtain a noise component image; inputting the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a noise-reduced RAW image; wherein, AI denoising network model comprises: a depth separable channel attention module, an inverted residual channel attention module, a short-circuited inverted residual channel attention module, a discrete wavelet transform module, and an inverse discrete wavelet transform module. The noise model is adopted to extract noise components, the discrete wavelet transform DWT and the inverse discrete wavelet transform IWT are combined to realize size compression and recovery of the characteristic diagram, point separable convolution, depth separable convolution, pre-activation, channel attention mechanism and residual structure are realized, a lightweight noise reduction network is realized, and the noise reduction effect is improved.

Description

Image noise reduction method and device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of image processing technologies, and in particular, to an image denoising method and apparatus, an electronic device, and a storage medium.

Background

In each imaging system, noise is inevitable due to the influence of sensors, illumination intensity, and the like, so that the image loses detail information and even becomes blurred. Especially in low light environments, image noise may seriously degrade image quality, and thus it is necessary to reduce the noise of an image. The image denoising technology is used for extracting characteristics of a noise image, removing noise by means of image priori knowledge, image self-similarity, multi-frame image complementary information and the like, reserving image details and reconstructing a high-quality image, and has important application value in the fields of mobile phone photographing videos, high-definition televisions, monitoring equipment, satellite images, medical images and the like.

Image noise reduction algorithms are mainly classified into conventional filtering methods and learning-based methods. The current mainstream traditional noise reduction algorithm is divided into two types: firstly, a Block-matching and 3D filtering (BM 3D) algorithm is adopted, but for scenes with severe noise, the noise still cannot be removed well, and discontinuous noise points are easily generated; and secondly, multi-frame images are superposed, the multi-frame images are matched through alignment, random noise is eliminated through multi-frame complementary information, but a ghost image area appears in the images possibly, and the method is only suitable for a photographing scene and cannot be suitable for a video scene with high real-time requirement.

With the development of Artificial Intelligence (AI) technology, an AI noise reduction algorithm using data driving is developed. The AI denoising algorithm realizes image denoising by learning the rule of large-scale noise image-clean image pair, and the denoising performance of the AI denoising algorithm is higher than that of the traditional denoising algorithm. However, in the prior art, the scale and the operation amount of the AI noise reduction algorithm model are large, or the model is small but the noise reduction effect is not ideal.

Disclosure of Invention

The present invention provides an image denoising method, apparatus, electronic device and storage medium, which can at least partially solve the problems in the prior art.

In order to achieve the purpose, the invention adopts the following technical scheme:

in a first aspect, an image denoising method is provided, including:

acquiring a target RAW map;

processing the target RAW image by using a pre-acquired noise model to obtain a noise component image;

inputting the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a RAW image after noise reduction;

wherein the AI denoising network model comprises: a depth separable channel attention module, an inverted residual channel attention module, a short-circuited inverted residual channel attention module, a discrete wavelet transform module, and an inverse discrete wavelet transform module.

Further, the number of the depth-separable channel attention modules is 6, which are respectively the first depth-separable channel attention module to the sixth depth-separable channel attention module; the number of the inverted residual channel attention modules is 3, and the inverted residual channel attention modules are respectively a first inverted residual channel attention module to a third inverted residual channel attention module; the number of the short-circuit inversion residual error channel attention modules is 3, and the short-circuit inversion residual error channel attention modules are respectively a first short-circuit inversion residual error channel attention module to a third short-circuit inversion residual error channel attention module; the number of the discrete wavelet transform modules is 2, and the discrete wavelet transform modules are respectively a first discrete wavelet transform module and a second discrete wavelet transform module; the number of the inverse discrete wavelet transform modules is 2, and the inverse discrete wavelet transform modules are respectively a first inverse discrete wavelet transform module and a second inverse discrete wavelet transform module;

the AI denoising network model further includes: the device comprises a first splicing module, a second splicing module, an adding module and a 3x3 convolution module;

wherein the first discrete wavelet transform module, the first depth separable channel attention module, the first short-circuited inverted residual channel attention module, the second discrete wavelet transform module, the second depth separable channel attention module, the first inverted residual channel attention module, the third depth separable channel attention module, the second inverted residual channel attention module, the first stitching module, the fourth depth separable channel attention module, the third inversion residual channel attention module, the first inverse discrete wavelet transform module, the second splicing module, the fifth depth separable channel attention module, the second short-circuit inversion residual channel attention module, the second inverse discrete wavelet transform module, the sixth depth separable channel attention module, the third short-circuit inversion residual channel attention module, the 3x3 convolution module and the addition module are sequentially connected; the addition module is also accessed to a target RAW graph; the first short-circuit inversion residual error channel attention module is further connected with the second splicing module, and the output of the first inversion residual error channel attention module is further connected with the first splicing module.

Further, the depth separable channel attention module comprises: the device comprises a first band leakage correction linear unit, a first point convolution unit, a second band leakage correction linear unit, a first depth convolution unit, a first global average pooling unit, a first 1x1 point convolution unit, a first nonlinear activation unit, a second 1x1 point convolution unit, a first Sigmoid unit, a first multiplication unit and a second point convolution unit which are connected in sequence; wherein the first deep convolution unit is further connected with the first multiplication unit.

Further, the inverted residual channel attention module comprises: the third band leakage correction linear unit, the third point convolution unit, the fourth band leakage correction linear unit, the second depth convolution unit, the second global average pooling unit, the third 1x1 point convolution unit, the second nonlinear activation unit, the fourth 1x1 point convolution unit, the second Sigmoid unit, the second multiplication unit, the fourth point convolution unit and the first addition unit are connected in sequence; the second deep convolution unit is further connected with the second multiplication unit, and the first addition unit is further connected with the input of the third leakage correction linear unit.

Further, the short inversion residual channel attention module comprises: a 3x3 convolution unit, a fifth band leakage correction linear unit, a fifth point convolution unit, a sixth band leakage correction linear unit, a third depth convolution unit, a third global average pooling unit, a fifth 1x1 point convolution unit, a third nonlinear activation unit, a sixth 1x1 point convolution unit, a third Sigmoid unit, a third multiplication unit, a sixth point convolution unit and a second addition unit which are connected in sequence; the third deep convolution unit is further connected with the third multiplication unit, and the second addition unit is further connected with the input of the fifth band leakage correction linear unit through a 3x3 convolution unit.

Further, the image denoising method further comprises:

acquiring RAW image data sets of an imaging system to be denoised under different scenes;

preprocessing RAW images in RAW image data sets under different scenes to obtain noise level statistical information of the RAW images;

and respectively fitting the noise level statistical information of the RAW image corresponding to the RAW image data set in each scene to obtain a Poisson noise parameter curve and a Gaussian noise parameter curve which are related to ISO (International standardization organization) and serve as noise models for acquiring noise component images.

Further, the image denoising method further comprises:

and restoring the RAW image subjected to noise reduction to obtain an RGB image subjected to noise reduction.

Further, the recovery process includes: white balancing, demosaicing, color space transformation, tone transformation.

Further, before the processing the target RAW map by using the pre-acquired noise model to obtain a noise component image, the method further includes:

and carrying out RAW domain preprocessing on the target RAW map.

Further, the RAW domain preprocessing comprises: dark level correction, fixed pattern noise correction, dead-spot correction, Lens shaping correction, green balancing.

Further, the image denoising method further comprises:

simulating an ISP processing flow of an imaging system to be denoised, and converting a high-quality RGB image in an image database into a clean RAW image by using an inverse ISP;

adding Gaussian-Poisson noise of a random ISO level to the clean RAW image to obtain a noise-containing RAW image;

and training a pre-established AI noise reduction network model by using the noise-containing RAW map and the clean RAW map.

In a second aspect, there is provided an image noise reduction device comprising:

the data acquisition module is used for acquiring a target RAW image;

the noise component extraction module is used for processing the target RAW image by utilizing a pre-acquired noise model to obtain a noise component image;

the noise reduction module is used for inputting the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a RAW image after noise reduction;

In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the image noise reduction method when executing the program.

In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, realizes the steps of the image noise reduction method as described above.

The embodiment of the invention provides an image noise reduction method, an image noise reduction device, electronic equipment and a storage medium, wherein the image noise reduction method comprises the following steps: acquiring a target RAW map; processing the target RAW image by using a pre-acquired noise model to obtain a noise component image; inputting the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a RAW image after noise reduction; wherein the AI denoising network model comprises: a depth separable channel attention module, an inverted residual channel attention module, a short-circuited inverted residual channel attention module, a discrete wavelet transform module, and an inverse discrete wavelet transform module. The method comprises the steps of extracting noise components by adopting a noise model, realizing size compression and recovery of a feature map by combining Discrete Wavelet Transform (DWT) and inverse discrete wavelet transform (IWT), realizing point separable convolution, depth separable convolution, pre-activation, channel attention mechanism and residual structure by adopting a depth separable channel attention module, an inverted residual channel attention module and a short inverted residual channel attention module, and realizing a lightweight noise reduction network and improving a noise reduction effect by focusing on difference calculation of a noise RAW map and a clean RAW map and reducing calculated amount and model size.

In order to make the aforementioned and other objects, features and advantages of the invention comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts. In the drawings:

FIG. 1 is a first flowchart illustrating an image denoising method according to an embodiment of the present invention;

FIG. 2 illustrates an AI denoising network model in an embodiment of the invention;

FIG. 3 illustrates a specific structure of a depth separable channel attention module in an embodiment of the present invention;

FIG. 4 shows a specific structure of an inverted residual channel attention module in an embodiment of the present invention;

FIG. 5 shows a specific structure of a short inverted residual channel attention module in an embodiment of the present invention;

FIG. 6 is a second flowchart illustrating an image denoising method according to an embodiment of the present invention;

FIG. 7 illustrates an image denoising process in an embodiment of the present invention;

FIG. 8 shows the detailed steps of step S600 in an embodiment of the present invention;

FIG. 9 shows the shot noise figure log in an embodiment of the invention₁₀(λ_shot) And log₁₀(ISO) relation;

FIG. 10 is a Gaussian distribution histogram showing the difference value Δ in the embodiment of the present invention;

FIG. 11 shows the read noise factor ε (λ) in an embodiment of the present invention_read) And log₁₀(ISO) relation;

fig. 12 shows the detailed steps of step S700 in the embodiment of the present invention;

FIG. 13 illustrates a model training process in an embodiment of the invention;

FIG. 14 shows image processing effect comparison in an embodiment of the present invention;

fig. 15 is a block diagram of the structure of an image noise reduction apparatus in the embodiment of the present invention;

fig. 16 is a block diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

It should be noted that the terms "comprises" and "comprising," and any variations thereof, in the description and claims of this application and the above-described drawings, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.

The existing AI noise reduction algorithm needs a large-scale real noise image-clean image to train a data set, and for different imaging systems, because the noise mode of the imaging system is affected by differences of noise modes, such as color filter change, lens coating adjustment, image sensor change and the like. Therefore, each time an imaging system component is adjusted, a corresponding data set needs to be collected for model training, which becomes a heavy and unrealistic task; if the collection is carried out after the imaging system is mature and stable, the project development progress is possibly delayed. In the prior art, aiming at the same scene, a low ISO and long exposure image is shot as a clean image, a high ISO and short exposure image is shot as a noise image, and camera parameters such as exposure time and the like are adjusted to enable the brightness of the two images to be consistent; the method has the disadvantages that the low ISO and long exposure image is taken as a clean image, and the residual noise is inevitable; and the data acquisition workload is large, moving objects and luminance illumination change cannot exist in a shooting scene, and the shooting scene is limited by the scene, ISO and exposure time. In the embodiment of the invention, the noise-containing RAW Image is synthesized by simulating the processing process of an Image Signal Processor (ISP) and the noise model of the RAW Image, and the Image pair is formed by the noise-free RAW Image, so that the neural network noise reduction model training is carried out, and the technical problem is effectively solved.

The AI denoising technology has a large difference, and in order to meet the AI denoising requirements of a mobile terminal, such as low calculation amount, high performance, and low storage space, a lightweight design needs to be performed on the AI denoising network. The embodiment of the invention provides a feasible lightweight AI noise reduction network model, which only uses a single-frame RAW image to realize noise reduction, can be used for image and video noise reduction, and reduces the cache requirement in the image processing process.

Due to the improvement of the imaging sensor technology, an additive white Gaussian noise component independent from a signal becomes weaker and weaker; and the signal dependent gaussian-poisson noise component becomes more and more important. Therefore, the embodiment of the invention mainly aims at the noise model definition and the noise reduction model training of Gaussian-Poisson noise.

In the noise estimation method based on uniform image areas in the prior art, the image is required to have certain image areas without textures and with few textures, the uniform image areas are segmented through edge detection, and then noise variance estimation is carried out; the method cannot be applied to scenes with rich textures, and is poor in robustness. In the noise estimation method based on image filtering, image edges and noise are separated through a high-pass filter and a low-pass filter, but the image edges and the noise cannot be completely separated, so that the estimated noise level is greatly different from the actual noise level. The noise estimation method based on wavelet transform separates out high frequency components through multiple times of wavelet transform, removes wavelet coefficients higher than a threshold value in the high frequency components and realizes noise reduction; the high-frequency wavelet coefficient of the method is still influenced by the image structure and texture, and the accuracy of noise estimation is also influenced by the threshold value selection algorithm, so that the accuracy and robustness of the whole noise estimation are still low. The embodiment of the invention mainly adopts a local noise estimation method based on image blocks, and generates estimation parameters of a Gaussian-Poisson noise model by calculating the noise parameters of the image blocks, the characteristic values of the covariance matrix and the statistical information of the characteristic values of the image blocks.

The embodiment of the invention relates to the technical field of image processing, and the method adopts a local noise estimation method based on an image block to generate a noise-containing RAW image and a noise-free RAW image from a high-quality RGB image, so that the synthesized noise-containing RAW image is similar to a real noise RAW image, an AI noise reduction network is trained to meet the single-frame noise reduction requirement of the real image and a video, and the collection work of a heavy noise image-clean image on a training data set is avoided. In addition, the embodiment of the invention carries out noise estimation and training based on a Gaussian-Poisson noise model (PG noise). The model assumes that the original RAW map noise is mainly composed of signal-dependent shot noise and signal-independent read noise. The shot noise is approximately Poisson distribution and is caused by photon arrival statistical error; the read noise is approximately gaussian distributed and is caused by inaccuracies in the read-out circuitry.

Fig. 1 is a first flowchart illustrating an image denoising method in an embodiment of the present invention, where the image denoising method is applied to a server or an electronic terminal device, for example: smart phones, tablet electronic devices, network set-top boxes, portable computers, desktop computers, Personal Digital Assistants (PDAs), vehicle-mounted devices, smart wearable devices, toys, smart home control devices, pipeline device controllers, monitors, and the like. Wherein, intelligence wearing equipment can include intelligent glasses, intelligent wrist-watch, intelligent bracelet etc..

As shown in fig. 1, the image denoising method may include the following:

step S100: acquiring a target RAW map;

the target RAW map is a noise-containing RAW map photographed by an imaging system to be noise-reduced (such as a camera, a mobile phone, a video camera, a monitor).

Step S200: processing the target RAW image by using a pre-acquired noise model to obtain a noise component image;

in particular, the noise model may be a gaussian-poisson noise model.

Step S300: inputting the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a RAW image after noise reduction;

Specifically, the AI noise reduction network model provided in the embodiment of the present invention adopts a depth separable Channel Attention Block (DCAB), an Inverted Residual Channel Attention Block (IRCAB), a short-circuited Inverted Residual Channel Attention Block (SIRCAB) to replace a conventional convolutional layer and an activation function, and takes into account multi-scale feature information while reducing the amount of computation and the size of the model; and replacing operations such as pooling, down-sampling, up-sampling and deconvolution of the feature layer by Discrete Wavelet Transform (DWT) and Inverse Discrete Wavelet Transform (IWT), so as to ensure that information is not lost.

The image noise reduction method provided by the embodiment of the invention adopts the depth separable channel attention module, the inverted residual channel attention module and the short inverted residual channel attention module to realize point separable convolution, depth separable convolution, pre-activation, a channel attention mechanism and a residual structure, and aims at the difference calculation of a noise RAW image and a clean RAW image and reducing the calculated amount and the size of a model, so that a lightweight noise reduction network is realized, and the noise reduction effect is improved.

In an alternative embodiment, referring to fig. 2, the number of depth-separable channel attention modules is 6, first through sixth depth-separable channel attention modules DCAB1 through DCAB6, respectively; the number of the inverted residual channel attention modules is 3, and the inverted residual channel attention modules are respectively a first inverted residual channel attention module IRCAB1 to a third inverted residual channel attention module IRCAB 3; the number of the short-circuit inversion residual error channel attention modules is 3, and the short-circuit inversion residual error channel attention modules are respectively a first short-circuit inversion residual error channel attention module SIRCAB1 to a third short-circuit inversion residual error channel attention module SIRCAB 3; the number of the discrete wavelet transform modules is 2, and the discrete wavelet transform modules are respectively a first discrete wavelet transform module DWT1 and a second discrete wavelet transform module DWT 2; the number of the inverse discrete wavelet transform modules is 2, and the inverse discrete wavelet transform modules are respectively a first inverse discrete wavelet transform module IWT1 and a second inverse discrete wavelet transform module IWT 2;

the AI denoising network model further includes: a first concatenation module C1, a second concatenation module C2, a summing module D1, and a 3x3 convolution module 3x3 conv;

wherein the first discrete wavelet transform module DWT1, the first depth separable channel attention module DCAB1, the first shorted inverted residual channel attention module SIRCAB1, the second discrete wavelet transform module DWT2, the second depth separable channel attention module DCAB2, the first inverted residual channel attention module IRCAB1, the third depth separable channel attention module DCAB3, the second inverted residual channel attention module IRCAB2, the first stitching module C1, the fourth depth separable channel attention module DCAB4, the third inverted residual channel attention module IRCAB3, the first inverted discrete wavelet transform module IWT1, the second stitching module C2, the fifth depth separable channel attention module AB5, the second shorted residual channel attention module SIRCAB2, the second discrete inverse wavelet transform module IWT2, the sixth depth channel attention module DCT 6, the third depth separable channel attention module SIRCAB3, the shorted residual channel attention module SIRCAB 35, The 3x3 convolution module 3x3conv and the addition module D1 are connected in sequence; the addition module is also accessed to a target RAW map Noisy RAW; the first short inverted residual channel attention module SIRCAB1 is also connected to the second stitching module C2, and the output of the first inverted residual channel attention module IRCAB1 is also connected to the first stitching module C1 block.

Here, noise Level represents a noise component image, and Denoised RAW represents a RAW image after noise reduction.

It is worth noting that the AI noise reduction net parameter is 0.96M, and for 4K real-shot video, the real-shot RAW graph noise reduction calculation amount of 2160 × 4096 is about 69.66Gmacs, which is greatly reduced compared with the prior art.

Discrete Wavelet Transform (DWT) can separate the characteristic layer into high-frequency, intermediate-frequency and low-frequency channels, and a subsequent network behind a DWT operator can reduce noise of the characteristic layer; the inverse discrete wavelet transform can combine high-frequency, intermediate-frequency and low-frequency characteristic layers, and improve the image processing precision.

In an alternative embodiment, referring to fig. 3, depth separable channel attention module DCAB includes: a first leakage correction linear unit LeakyReLU1, a first point convolution unit pw-conv1, a second leakage correction linear unit LeakyReLU2, a first depth convolution unit dw-conv1, a first global average pooling unit AvgPool1, a first 1x1 point convolution unit 1x1 conv1, a first nonlinear activation unit ReLU1, a second 1x1 point convolution unit 1x1 conv2, a first Sigmoid unit S1, a first multiplication unit T1 and a second point convolution unit pw-conv2 which are connected in sequence; wherein the first deep convolution unit dw-conv1 is further connected to the first multiplication unit T1.

In an alternative embodiment, referring to fig. 4, the inverted residual channel attention module IRCAB includes: a third band leakage correction linear unit LeakyReLU3, a third point convolution unit pw-conv3, a fourth band leakage correction linear unit LeakyReLU4, a second depth convolution unit dw-conv2, a second global average pooling unit AvgPool2, a third 1x1 point convolution unit 1x1 conv3, a second nonlinear activation unit ReLU2, a fourth 1x1 point convolution unit 1x1 conv4, a second Sigmoid unit S2, a second multiplication unit T2, a fourth point convolution unit pw-conv4 and a first addition unit D2 which are connected in sequence; the second depth convolution unit dw-conv2 is further connected to the second multiplication unit T2, and the first addition unit D1 is further connected to the input of the third leakage correction linear unit leak relu 3.

In an alternative embodiment, referring to fig. 5, the short inversion residual channel attention module SIRCAB comprises: a 3x3 convolution unit 3x3conv, a fifth leakage correction linear unit LeakyReLU5, a fifth point convolution unit pw-conv5, a sixth leakage correction linear unit LeakyReLU6, a third depth convolution unit dw-conv3, a third global average pooling unit AvgPool3, a fifth 1x1 point convolution unit 1x1 conv5, a third nonlinear activation unit ReLU3, a sixth 1x1 point convolution unit 1x1 conv3, a third Sigmoid unit S3, a third multiplication unit T3, a sixth point convolution unit pw-conv6 and a second addition unit D3 which are connected in sequence; wherein, the third depth convolution unit dw-conv3 is further connected with the third multiplication unit T3, and the second addition unit D3 is further connected with the input of the fifth band leakage correction linear unit leak relu5 through a 3x3 convolution unit.

Specifically, the LeakyReLU is set at the input of each of the DCAB, the IRCAB and the SIRCAB, so that pre-activation before point convolution and depth convolution is realized, in addition, the SIRCAB short-circuits the input to the addition unit through 3x3conv, namely, the network model provided by the embodiment of the invention obtains the depth separable channel attention module DCAB, the inverted residual channel attention module IRCAB and the short-circuited inverted residual channel attention module SIRCAB by respectively applying pre-activation and short-circuit on the basis of the inverted residual attention module of the MobileNetV 3.

In addition, in the depth separable channel attention module DCAB, the inverted residual channel attention module IRCAB and the short inverted residual channel attention module SIRCAB, the channel attention mechanism is realized by the cooperation of AvgPool, 1x1 conv, RELU, 1x1 conv, Sigmoid and the multiplication unit. In the IRCAB module, the addition unit is also connected to the input of the leakyreu, forming a residual structure.

For example, the noise-containing RAW graph and the noise component image of h × w × 1 size are respectively normalized into two h/2 × w/2 × 4 size data by four channels of R, Gr, Gb, and B as AI network inputs. Through the first DWT, 4 frequency components of high, middle and low are extracted and combined in a lossless manner, and the dimension is compressed into h/4 w/4, namely 1/2 w/1/2 feature layers; compressing the dimension into h/8 w/8 through a second DWT to obtain 1/4 w 1/4 feature layers; through the first IWT, the dimension is restored to h/4 w/4, namely 1/2 w 1/2 feature layers; through a second IWT, the dimension is restored to h/2 w/2, namely a 1 w 1 feature layer; through a 1x1 convolution kernel and a 3x3 convolution kernel, information extraction and interaction among 2 x 2, 6 x 6, 4 x 4, 12 x 12, 8 x 8 and 24 x 24 area pixels on the original RAW graph can be realized respectively, so that noise introduced by ISP back-end image processing is reduced; and information extraction and interaction among channels are realized through a channel attention mechanism, and feature channels of different layers are focused. Based on the splicing modules C1 and C2, the front and rear feature layers are spliced, so that the transmission efficiency of the feature layers is guaranteed, the training stability is improved, and the convergence is easier. The front and back combination form of DCAB, SIRCAB and IRCAB can avoid the direct transmission of characteristic information through a direct connection branch and the failure of a residual error branch. And finally, adding the noise-containing RAW map and the noise-containing RAW map to obtain a noise-reduced clean RAW map, so that the intermediate network mainly focuses on the calculation of the difference between the noise-containing RAW map and the clean RAW map, and the interference of the noise-containing RAW maps with different scenes and textures on the noise reduction performance is avoided.

Where pw _ conv indicates pointwise contribution, and dw _ conv indicates depthwise contribution. It should be noted that the AI noise reduction network in the embodiment of the present invention is not limited to this, and other CNN, MLP, and Transformer networks may be considered to implement similar functions in the embodiment of the present invention.

In an alternative embodiment, referring to fig. 6, the image denoising method may further include the following:

step S500: and restoring the RAW image subjected to noise reduction to obtain an RGB image subjected to noise reduction.

Specifically, the recovery processing includes: white balancing, demosaicing, color space transformation, tone transformation.

By adopting the technical scheme, the precision and the effect of image processing can be improved.

In an alternative embodiment, with continued reference to fig. 6, the image denoising method may further include the following:

step S400: and carrying out RAW domain preprocessing on the target RAW map.

Specifically, RAW domain preprocessing includes: dark level correction, fixed pattern noise correction, dead-spot correction, Lens shaping correction, green balancing.

FIG. 7 illustrates an image denoising process in an embodiment of the present invention; as shown in fig. 7, during the application inference phase: the method comprises the steps that an imaging system (an image sensor) to be denoised shoots a RAW image containing noise, RAW domain preprocessing is carried out on the RAW image containing the noise, wherein the RAW domain preprocessing comprises one or more of dark level correction, fixed mode noise correction, dead pixel correction, Lens Shading correction and green balance, then the preprocessed RAW image containing the noise is input into a Gaussian-Poisson noise model, a noise parameter curve is applied, shooting parameters (such as ISO, shooting scene, camera model and the like) are used as prior information, corresponding Poisson noise parameters and Gaussian noise parameters are obtained, and a noise component image is obtained by combining the RAW image containing the noise; splicing the noise component image and the RAW image containing the noise, inputting the image into an AI noise reduction network for noise reduction, and obtaining a clean RAW image after noise reduction; if necessary, the image can be transmitted to a back-end processing chip for white balance, demosaicing, color space conversion, tone conversion and the like, and a clean RGB image is recovered.

When splicing, the two 4-channel images are spliced into an 8-channel image, which may be that the red channel of the noise component image, the red channel of the noise-containing RAW image, the green channel of the noise component image and the green channel of the noise-containing RAW image are spliced together, or the red, green and blue of the noise component image and the red, green and blue of the noise-containing RAW image are spliced in sequence.

In addition, the scene information may be acquired through various methods. Dividing the images into day shooting and night shooting according to the shooting time and the time zone; through scene detection, images are divided into outdoor shooting, indoor shooting, figure shooting, landscape shooting, article shooting and the like; dividing the image into a bright scene and a dark scene through image illumination analysis; and dividing the image into whether to shoot in a backlight mode or not through image contrast and dynamic range analysis.

step S600: acquiring a noise model;

step S700: and training the AI denoising network model.

Those skilled in the art can understand that the steps S600 and S700 are completed before the inference phase is applied, and the implementation time, sequence, etc. of the steps S600 and S700 can be freely arranged according to the project requirement, which is not limited in the embodiment of the present invention.

FIG. 8 shows the detailed steps of step S600 in an embodiment of the present invention; as shown in fig. 8, this step S600 may include the following:

step S610: acquiring RAW image data sets of an imaging system to be denoised under different scenes;

step S620: preprocessing RAW images in RAW image data sets under different scenes to obtain noise level statistical information of the RAW images;

step S630: and respectively fitting the noise level statistical information of the RAW image corresponding to the RAW image data set in each scene to obtain a Poisson noise parameter curve and a Gaussian noise parameter curve which are related to ISO (International standardization organization) and serve as noise models for acquiring noise component images.

Specifically, the method includes the steps of collecting RAW image data sets of an imaging system to be denoised in different scenes, respectively obtaining denoising models for the scenes, and preprocessing the RAW images for the scenes, wherein the preprocessing includes any one or more of the following steps: dark level correction, Fixed Pattern Noise (FPN) correction, dead pixel correction, Lens Shading correction, green balance. For the processed RAW map data set, an image block based local noise estimation method is used,and acquiring noise level statistical information of the RAW image, fitting to acquire a Poisson noise parameter curve and a Gaussian noise parameter curve related to ISO (International organization for standardization), taking the Poisson noise parameter curve and the Gaussian noise parameter curve as a noise model, and pre-storing the noise model into a memory. In particular, the Gaussian-Poisson noise model x_n＝y_n+ε_n(y_n)，

Wherein the unknown clean image is y_nThe true noisy image is x_n(ii) a Signal dependent poisson-gaussian noise of epsilon_nN represents pixels, each pixel has a noise offset conforming to a normal distribution N with a standard deviation of

The noise estimation method in the embodiment of the invention obtains shot noise parameter lambda according to the statistical information_shotCurve, read noise parameter lambda_readCurve, and log₁₀(ISO) satisfies a quadratic polynomial relationship, each having a coefficient a₀/a₁/a₂、b₀/b₁/b₂. Read noise factor ε (λ)_read). Comprises the following steps:

x_n＝min(max(y_n+ε_n(y_n) 0),1) (in practice, x is required_n、y_nSatisfies the range of [0,1 ]])

log₁₀(λ_shot)＝a₀+a₁·log₁₀(ISO)+a₂·log₁₀(ISO)²

ε(λ_read)＝b₀+b₁·log₁₀(ISO)+b₂·log₁₀(ISO)²

λ_read＝ε(λ_read)·ε(λ_read)·λ_shot

The embodiment of the invention mainly adopts a local noise estimation method based on image blocks, and generates estimation parameters of a Gaussian-Poisson noise model by calculating the noise parameters of the image blocks, the characteristic values of the covariance matrix and the statistical information of the characteristic values of the image blocks. The method specifically comprises the following steps:

(1) the RAW map was normalized to [0,1 ] by the above-described pretreatment]Is divided into several sizes s₁*s₂The image blocks are subjected to color channel separation according to R, Gr, Gb and B channels and are normalized into 4 s₁/2*s₂For example, it may take s1 and s2 as 512 respectively, and perform block calculation as a large image block to improve the calculation efficiency;

(2) extracting the size d for each large image block₁*d₂Step ds, number of channels 4, e.g. d may be taken₁、d₂16 respectively, and ds is 7 as a small image sequence set; each large image block has c_nSmall image blocks;

(3) calculating a covariance matrix of each small image block and a characteristic value of the matrix; counting and analyzing the characteristic values of all small image blocks in the large image block, and solving the root number of the median of the characteristic value sequence to be used as the shot noise coefficient of the large image block; besides the median, effective information of the characteristic value sequence can be extracted by selecting the modes of average number, weighted average and the like to be used as the shot noise coefficient of the large image block;

(4) calculating the shot noise coefficient of each large image block, and averaging to obtain the shot noise coefficient of the RAW image;

(5) statistically analyzing the relation between the shot noise coefficient and ISO of each RAW image, and obtaining the log through least square fitting₁₀(λ_shot)＝a₀+a₁·log₁₀(ISO)+a₂·log₁₀(ISO)²+N(μ＝0,σ_s) Actual data are distributed on two sides of the curve, and the offset meets normal distribution with the expectation of 0 and the standard deviation of sigma s; FIG. 9 shows the shot noise figure log in an embodiment of the invention₁₀(λ_shot) And log₁₀(ISO) relationship curve.

(6) The normal distribution N in the step (5) is obtained by analyzing a Gaussian distribution histogram of the difference value delta between the shot noise distribution and a quadratic fit curve thereof and is used as the shot noise log in the training stage₁₀(λ_shot) The curve Gaussian random component enables the training noise to be closer to the real noise distribution; referring to FIG. 10, there is shown an implementation of the present inventionA gaussian distribution histogram of the difference value Δ in the example;

(7) analyzing the difference values of Delta and log in the step (6)₁₀(ISO) relationship, obtaining a quadratic fit curve epsilon (lambda) by a least square method_read)＝b₀+b₁·log₁₀(ISO)+b₂·log₁₀(ISO)²+N(μ＝0,σ_r) As the read noise factor ε (λ)_read) Curves in relation to ISO; wherein, the actual data are distributed on both sides of the curve, mu represents the expected value or the average value of normal distribution, and the offset meets the normal distribution that the expected mu is 0 and the standard deviation is sigma r; referring to FIG. 11, a read noise factor ε (λ) in an embodiment of the present invention is shown_read) And log₁₀(ISO) relation;

(8) step (7) reading noise factor epsilon (lambda) through analyzing normal distribution N_read) The Gaussian distribution histogram of the difference from its quadratic fit curve is obtained as the read noise factor ε (λ) of the training phase_read) The curve Gaussian random component enables the training noise to be closer to the real noise distribution;

(9) calculating a read noise parameter by reading a noise factor and a shot noise parameter

λ_read＝ε(λ_read)·ε(λ_read)·λ_shot

Fig. 12 shows the detailed steps of step S700 in the embodiment of the present invention; as shown in fig. 12, this step S700 may include the following:

step S710: simulating an ISP processing flow of an imaging system to be denoised, and converting a high-quality RGB image in an image database into a clean RAW image by using an inverse ISP;

step S720: adding Gaussian-Poisson noise of a random ISO level to the clean RAW image to obtain a noise-containing RAW image;

step S730: and training a pre-established AI noise reduction network model by using the noise-containing RAW map and the clean RAW map.

Specifically, the training phase: and simulating an ISP processing flow of the imaging system to be denoised, and converting the high-quality RGB image in the existing image database into a clean RAW image by using an inverse ISP. Generating ISO randomly, and inputting the ISO into a noise model to obtain Gaussian-Poisson noise; this ISO level gaussian-poisson noise is added to the clean RAW map, and a noisy RAW map is synthesized. And using the synthesized noise-containing RAW image-clean RAW image pair for the AI denoising network training.

In the embodiment of the invention, through inverse ISP conversion and a noise model, the numerical distribution of the R, Gr, Gb and B four-channel images containing noise synthesized from the RGB images in the same scene is ensured to be close to the numerical distribution of the RAW image shot by an imaging system to be denoised, and the AI denoising network is ensured to still have better denoising performance on the actually shot RAW image.

Description of training phase, see fig. 13:

(1) inputting a high-quality RGB image, and performing inverse ISP transformation (including inverse tone mapping, inverse Gamma correction, inverse color space transformation, inverse AWB correction, and inverse demosaicing) to obtain a clean RAW image y_nRandomly selecting an ISO according to the Poisson noise figure log₁₀(λ_shot) And gaussian noise factor epsilon (lambda)_read) And log₁₀(ISO) quadratic fitting curve, calculating to obtain Poisson noise parameter lambda corresponding to ISO_shotAnd Gaussian noise parameter λ_read；

(2) According to the relation of noise formula and clean RAW diagram

Obtaining a noise value related to each pixel and a signal, and adding the noise value to a clean RAW map to obtain a noise RAW map, so that the generation process of the noise RAW map contained in the imaging system is simulated;

(3) calculating formula xi according to noise component image_n(x_n)＝λ_shot·x_n+λ_readObtaining a noise component image xi of the noise RAW map_n(x_n) After being spliced with the noise RAW image, the noise-reduced RAW image is input into an AI noise-reduction network for noise reduction training to obtain a noise-reduced RAW image; and comparing the clean RAW image serving as a label image with the noise-reduced RAW image, and optimizing the AI noise-reduced network by using L1 Loss.

Applying inference phase specification:

(1) calculating a corresponding ISO value according to the digital gain and the analog gain of the image sensor (serving as image data along with the image);

(2) according to Poisson noise coefficient log₁₀(λ_shot) And gaussian noise factor epsilon (lambda)_read) And log₁₀(ISO) quadratic fitting curve to obtain Poisson noise parameter lambda_shotGaussian noise parameter lambda_read；

log₁₀(λ_shot)＝a₀+a₁·log₁₀(ISO)+a₂·log₁₀(ISO)²

ε(λ_read)＝b₀+b₁·log₁₀(ISO)+b₂·log₁₀(ISO)²

λ_read＝ε(λ_read)·ε(λ_read)·λ_shot

It is worth noting that there are different noise models and noise reduction network models for different application scenarios. After the target image is acquired, the scene is identified according to image time, or identified through image analysis, or identified according to light intensity, or identified according to contrast. And selecting a corresponding noise model and a noise reduction network model according to the identified scene to perform noise reduction processing.

(3) Calculating formula xi according to noise component image_n(x_n)＝λ_shot·x_n+λ_readGenerating a noise component image ξ of the noise-containing RAW map_n(x_n) And after being spliced with the RAW graph containing the noise, the RAW graph is input into a neural network for noise reduction.

By adopting the technical scheme, the performance evaluation is carried out by using a DND (Darmstadt Noise dataset) data set according to the Noise level estimation method, the lightweight AI denoising network and the inverse ISP transformation. Table 1 is the noise reduction algorithm performance evaluation given by the DND data set official website. Among them, Peak Signal to Noise Ratio (PSNR) and Structural Similarity (SSIM) are two important evaluation indexes for image quality recovery. The contrast maps shown in fig. 14 are all restored to RGB maps using the same back-end image processing steps after the RAW domain noise reduction, and the image processing effect is shown in a to e in fig. 14.

TABLE 1

The embodiment of the invention designs a lightweight image denoising network, which adopts Discrete Wavelet Transform (DWT) and Inverse Discrete Wavelet Transform (IDWT) to realize size compression and recovery of a characteristic graph, adopts point separable convolution, depth separable convolution, pre-activation, a channel attention mechanism and a residual structure to pay attention to difference calculation of a noise RAW graph and a clean RAW graph and reduce the calculated amount and the model size, and constructs an effective lightweight denoising network. In the training process, the influence of the subsequent ISP flow on the noise mode is considered, a plurality of noise reduction processes distributed in the traditional ISP flow are combined into one AI noise reduction network, and the design complexity of the AI _ ISP is reduced. The single-frame RAW image is adopted to realize noise reduction, and the method can be used for noise reduction of images and video data.

In addition, the embodiment of the invention also provides a shot noise and read noise parameter estimation method of the Poisson-Gaussian noise model, which is characterized in that quadratic curves of shot noise, read noise and ISO are fitted through the characteristic values of the image block noise parameters and covariance matrix and the statistical information of the characteristic values of the image block, and the difference between the fitted curves and the real noise is considered, so that an AI noise reduction network under the training of a synthetic data set can be used for reducing the noise of a RAW (RAW image) of the real noise.

The method for estimating and defining the Gaussian-Poisson noise model can obtain a noise component image of a noise RAW image at a mobile terminal according to a predefined noise model, ISO and scene information, and reduce noise through an AI noise reduction network after splicing the noise component image with the noise RAW image; the calculation amount and the running time of blind noise estimation of a RAW graph containing noise at a mobile terminal are greatly reduced;

the noise-reduced RAW Image may be output to a back-end processing Chip to recover the RGB Image, where the back-end processing Chip may be a System on Chip (SoC) or an Image Signal Processor (ISP).

Based on the same inventive concept, the present application further provides an image noise reduction apparatus, which can be used to implement the methods described in the foregoing embodiments, as described in the following embodiments. The principle of solving the problem by the image noise reduction device is similar to the method, so the implementation of the image noise reduction device can refer to the implementation of the method, and repeated details are not repeated. As used hereinafter, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.

Fig. 15 is a block diagram of the structure of an image noise reduction apparatus in the embodiment of the present invention, specifically, including: the device comprises a data acquisition module, a noise component extraction module and a noise reduction module.

The data acquisition module is used for acquiring a target RAW map;

the noise component extraction module processes the target RAW image by using a pre-acquired noise model to obtain a noise component image;

the noise reduction module inputs the target RAW image and the noise component image into a pre-trained AI noise reduction network model to obtain a noise-reduced RAW image;

The image noise reduction device provided by the embodiment of the invention adopts the depth separable channel attention module, the inverted residual channel attention module and the short inverted residual channel attention module to realize point separable convolution, depth separable convolution, pre-activation, a channel attention mechanism and a residual structure, and aims at the difference calculation of a noise RAW image and a clean RAW image and reducing the calculated amount and the size of a model, so that a lightweight noise reduction network is realized, and the noise reduction effect is improved.

The embodiment of the invention also provides a chip which comprises an RAW data digital circuit processing module, a burnt neural network noise reduction model and a data input and output module; when the chip is executed, the noise reduction method of the RAW graph is realized.

The apparatuses, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or implemented by a product with certain functions. A typical implementation device is an electronic device, which may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

In a typical example, the electronic device specifically includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the steps of the image noise reduction method described above when executing the program.

Referring now to FIG. 16, shown is a schematic diagram of an electronic device 600 suitable for use in implementing embodiments of the present application.

As shown in fig. 16, the electronic apparatus 600 includes a Central Processing Unit (CPU)601 that can perform various appropriate works and processes according to a program stored in a Read Only Memory (ROM)602 or a program loaded from a storage section 608 into a Random Access Memory (RAM)) 603. In the RAM603, various programs and data necessary for the operation of the system 600 are also stored. The CPU601, ROM602, and RAM603 are connected to each other via a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.

The following components are connected to the I/O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 608 including a hard disk and the like; and a communication section 609 including a network interface card such as a LAN card, a modem, or the like. The communication section 609 performs communication processing via a network such as the internet. The driver 610 is also connected to the I/O interface 605 as needed. A removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 610 as necessary, so that a computer program read out therefrom is mounted as necessary on the storage section 608.

In particular, according to an embodiment of the present invention, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, an embodiment of the present invention includes a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the image noise reduction method described above.

In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 609, and/or installed from the removable medium 611.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the units may be implemented in one or more software and/or hardware when implementing the present application.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. An image noise reduction method, comprising:

acquiring a target RAW map;

2. The image denoising method according to claim 1, wherein the number of the depth separable channel attention modules is 6, respectively a first depth separable channel attention module to a sixth depth separable channel attention module; the number of the inverted residual channel attention modules is 3, and the inverted residual channel attention modules are respectively a first inverted residual channel attention module to a third inverted residual channel attention module; the number of the short-circuit inversion residual error channel attention modules is 3, and the short-circuit inversion residual error channel attention modules are respectively a first short-circuit inversion residual error channel attention module to a third short-circuit inversion residual error channel attention module; the number of the discrete wavelet transform modules is 2, and the discrete wavelet transform modules are respectively a first discrete wavelet transform module and a second discrete wavelet transform module; the number of the inverse discrete wavelet transform modules is 2, and the inverse discrete wavelet transform modules are respectively a first inverse discrete wavelet transform module and a second inverse discrete wavelet transform module;

3. The image denoising method of claim 1, wherein the depth separable channel attention module comprises: the device comprises a first band leakage correction linear unit, a first point convolution unit, a second band leakage correction linear unit, a first depth convolution unit, a first global average pooling unit, a first 1x1 point convolution unit, a first nonlinear activation unit, a second 1x1 point convolution unit, a first Sigmoid unit, a first multiplication unit and a second point convolution unit which are connected in sequence; wherein the first deep convolution unit is further connected with the first multiplication unit.

4. The image denoising method of claim 1, wherein the inverse residual channel attention module comprises: the third band leakage correction linear unit, the third point convolution unit, the fourth band leakage correction linear unit, the second depth convolution unit, the second global average pooling unit, the third 1x1 point convolution unit, the second nonlinear activation unit, the fourth 1x1 point convolution unit, the second Sigmoid unit, the second multiplication unit, the fourth point convolution unit and the first addition unit are connected in sequence; the second deep convolution unit is further connected with the second multiplication unit, and the first addition unit is further connected with the input of the third leakage correction linear unit.

5. The image denoising method of claim 1, wherein the short inversion residual channel attention module comprises: a 3x3 convolution unit, a fifth band leakage correction linear unit, a fifth point convolution unit, a sixth band leakage correction linear unit, a third depth convolution unit, a third global average pooling unit, a fifth 1x1 point convolution unit, a third nonlinear activation unit, a sixth 1x1 point convolution unit, a third Sigmoid unit, a third multiplication unit, a sixth point convolution unit and a second addition unit which are connected in sequence; the third deep convolution unit is further connected with the third multiplication unit, and the second addition unit is further connected with the input of the fifth band leakage correction linear unit through a 3x3 convolution unit.

6. The image noise reduction method according to claim 1, further comprising:

7. The image noise reduction method according to claim 1, further comprising:

8. The image noise reduction method according to claim 7, wherein the restoration processing includes: white balancing, demosaicing, color space transformation, tone transformation.

9. The image denoising method according to claim 1, wherein before the processing the target RAW map by using the pre-obtained noise model to obtain a noise component image, the method further comprises:

and carrying out RAW domain preprocessing on the target RAW map.

10. The image denoising method of claim 9, wherein the RAW domain preprocessing comprises: dark level correction, fixed pattern noise correction, dead-spot correction, Lens shaping correction, green balancing.

11. The image noise reduction method according to claim 1, further comprising:

12. An image noise reduction apparatus, comprising:

the data acquisition module is used for acquiring a target RAW image;

13. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the image denoising method according to any one of claims 1 to 11 are implemented when the program is executed by the processor.

14. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the image denoising method according to any one of claims 1 to 11.