Noise model estimation method and device, image processing method and device

Abstract

The present application relates to a noise model estimation method, device, computer equipment, storage medium and computer program product. The method comprises: obtaining a plurality of first RAW images captured by a camera at different gain values ​​in a dark environment, and determining a read noise model according to the plurality of first RAW images; in a light environment, outputting the read noise of each second RAW image through the read noise model; each second RAW image is captured by the camera at different gain values; determining a shot noise model based on each second RAW image and the read noise corresponding to each second RAW image; and obtaining a target noise model of the RAW image captured by the camera based on the read noise model and the shot noise model. The use of this method can improve the accuracy of the estimated noise model.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a noise model estimation method, an image processing method, a noise model estimation device, an image processing device, an electronic device, and a computer-readable storage medium. Background Art

[0002] In the image signal processing (ISP) system, the image input information is first stored in the form of raw data, which is the most original information. Because of this advantage, noise reduction in the raw domain has received more and more attention. However, in order to achieve better noise reduction effects in the raw domain, it is often necessary to estimate the noise model (noise profile) of the raw data.

[0003] Traditional noise model estimation methods usually estimate the noise model from the perspective of different noise distribution models, which leads to the problem that the estimated noise model is inaccurate. Summary of the invention

[0004] Embodiments of the present application provide a noise model estimation method, an image processing method, a noise model estimation device, an image processing device, an electronic device, and a computer-readable storage medium, which can improve the accuracy of the estimated noise model.

[0005] In a first aspect, the present application provides a method for estimating a noise model. The method comprises:

[0006] Acquire a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images;

[0007] In a light environment, outputting the read noise of each second RAW image through the read noise model; each second RAW image is acquired by the camera with different gain values;

[0008] Determining a shot noise model based on each of the second RAW images and read noise corresponding to each of the second RAW images;

[0009] Based on the read noise model and the shot noise model, a target noise model of the RAW image acquired by the camera is obtained.

[0010] In a second aspect, the present application also provides a noise model estimation device. The device comprises:

[0011] A read noise model determination module, used to obtain a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images;

[0012] a shot noise model determination module, configured to output the read noise of each second RAW image through the read noise model in a light environment; each second RAW image is acquired by the camera with different gain values; and determine the shot noise model based on each second RAW image and the read noise corresponding to each second RAW image;

[0013] The target noise model determination module is used to obtain the target noise model of the RAW image acquired by the camera based on the read noise model and the shot noise model.

[0014] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0015] Acquire a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images;

[0016] In a light environment, outputting the read noise of each second RAW image through the read noise model; each second RAW image is acquired by the camera with different gain values;

[0017] Determining a shot noise model based on each of the second RAW images and read noise corresponding to each of the second RAW images;

[0018] Based on the read noise model and the shot noise model, a target noise model of the RAW image acquired by the camera is obtained.

[0019] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0020] Acquire a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images;

[0021] In a light environment, outputting the read noise of each second RAW image through the read noise model; each second RAW image is acquired by the camera with different gain values;

[0022] Determining a shot noise model based on each of the second RAW images and read noise corresponding to each of the second RAW images;

[0023] Based on the read noise model and the shot noise model, a target noise model of the RAW image acquired by the camera is obtained.

[0024] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0025] Acquire a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images;

[0026] In a light environment, outputting the read noise of each second RAW image through the read noise model; each second RAW image is acquired by the camera with different gain values;

[0027] Determining a shot noise model based on each of the second RAW images and read noise corresponding to each of the second RAW images;

[0028] Based on the read noise model and the shot noise model, a target noise model of the RAW image acquired by the camera is obtained.

[0029] The above-mentioned noise model estimation method, device, electronic device, computer-readable storage medium and computer program product, the electronic device obtains multiple first RAW images captured by the camera with different gain values ​​in a dark environment, and determines a read noise model based on the multiple first RAW images, and the objectively existing read noise generated by the camera in the dark environment can be obtained through the read noise model. In a light environment, the read noise of each second RAW image is first output through the read noise model, and then the shot noise model in the light environment can be determined based on each second RAW image and the read noise corresponding to each second RAW image. Then, based on the read noise model in the dark environment and the shot noise model in the light environment, the target noise model of the RAW image captured by the camera can be accurately estimated, and the noise of the RAW image captured by the camera can be more accurately estimated through the target noise model.

[0030] In a sixth aspect, the present application provides an image processing method. The method comprises:

[0031] Get the original RAW image captured by the camera;

[0032] The target noise of the original RAW image is determined by a target noise model, and the original RAW image is processed based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0033] In a seventh aspect, the present application further provides an image processing device. The device comprises:

[0034] The acquisition module is used to obtain the original RAW image captured by the camera;

[0035] A processing module is used to determine the target noise of the original RAW image through a target noise model, and process the original RAW image based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0036] In an eighth aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0037] Get the original RAW image captured by the camera;

[0038] The target noise of the original RAW image is determined by a target noise model, and the original RAW image is processed based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0039] In a ninth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0040] Get the original RAW image captured by the camera;

[0041] The target noise of the original RAW image is determined by a target noise model, and the original RAW image is processed based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0042] In a tenth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0043] Get the original RAW image captured by the camera;

[0044] The target noise of the original RAW image is determined by a target noise model, and the original RAW image is processed based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0045] The above-mentioned image processing method, device, electronic device, computer-readable storage medium and computer program product, the electronic device obtains the original RAW image captured by the camera; the target noise model obtained by the above-mentioned noise model estimation method can more accurately determine the target noise of the original RAW image, so that the original RAW image can be more accurately processed based on the more accurate target noise to obtain a clearer target RAW image. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 is a flow chart of a method for estimating a noise model in one embodiment;

[0048] Figure 2 A first RAW image captured by an electronic device in a dark environment in an embodiment;

[0049] Figure 3 An image of one channel of a first RAW image in one embodiment;

[0050] Figure 4 Schematic diagram of a curve showing the relationship between a gain value and a scale parameter of an R channel in one embodiment;

[0051] Figure 5 Schematic diagram of a curve showing the relationship between a gain value and a scale parameter of a GR channel in one embodiment;

[0052] Figure 6 Schematic diagram of a curve showing the relationship between a gain value and a scale parameter of a GB channel in one embodiment;

[0053] Figure 7 Schematic diagram of a curve showing the relationship between a gain value and a scale parameter of a B channel in one embodiment;

[0054] Figure 8 is a schematic diagram of a standard color card in one embodiment;

[0055] Fig. 9It is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of an R channel in one embodiment;

[0056] Fig.10 It is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a GR channel in one embodiment;

[0057] Fig.11 It is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a GB channel in one embodiment;

[0058] Fig.12 It is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a B channel in one embodiment;

[0059] Fig.13 is a flowchart of an image processing method in one embodiment;

[0060] Fig.14 is a schematic diagram of a target RAW image in one embodiment;

[0061] Fig.15 is a structural block diagram of a noise model estimation device in one embodiment;

[0062] Fig.16 is a structural block diagram of an image processing device in one embodiment;

[0063] Fig.17 FIG. 4 is a diagram showing the internal structure of an electronic device in one embodiment. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] In one embodiment, Figure 1 As shown, a method for estimating a noise model is provided. This embodiment uses the method applied to an electronic device as an example. The electronic device can be a terminal or a server. The above method can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be but not limited to various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0066] In this embodiment, the method includes the following steps:

[0067] Step 102 : acquiring a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determining a read noise model according to the plurality of first RAW images.

[0068] The electronic device includes a camera, and the camera includes components such as a lens, an image sensor, and a motor. The gain value is the magnification of the photosensitive element in the image sensor. The RAW image is the original data of the digital signal converted by the image sensor from the captured light signal. The first RAW image is the original image data collected by the camera in a dark environment. The read noise model is a model for the read noise of the output image.

[0069] It should be noted that the first RAW image is original image data that has not been subjected to any linear processing, and therefore in the processing flow (pipeline) of recording data in the camera, the function of removing bad pixels can be turned on.

[0070] It can be understood that in the electronic system of a camera or electronic device, the photon signal captured by the image sensor is converted into digital information of the output image, that is, in a light environment, the light signal is converted into an electrical signal. The noise generated in this process can be divided into read noise and shot noise.

[0071] During the propagation of photons, some photon noise will be introduced. However, when there is no photon input, the camera system will also generate read noise due to the influence of dark level fluctuations, digital-to-analog conversion, etc. In other words, the data of the first RAW image taken in a dark environment is not zero, but fluctuates around the dark level. Read noise is the objective noise generated by the camera in a dark environment. Shot noise is the noise caused by the unevenness of electron emission in active devices (such as vacuum tubes) in communication equipment. The read noise model is a model for the read noise of the output image.

[0072] Optionally, the electronic device attaches the camera lens to light-absorbing paper and places it in a completely dark environment, controls the camera to capture multiple first RAW images with different gain values, and exports multiple first RAW images. It can be understood that the gain value is the magnification of the photosensitive element and has a great relationship with the output first RAW image. Therefore, different gain values ​​will directly affect the noise value. It is necessary to capture multiple first RAW images with different gain values ​​to obtain more comprehensive and accurate data.

[0073] Optionally, the electronic device can also control the camera to capture multiple first RAW images with different gain values ​​and different exposure times. The exposure time is used as a random dynamic parameter. Different exposure times will cause the camera to heat up and affect the noise level to a certain extent. However, in order to collect comprehensive data, it is necessary to capture the first RAW images with different exposure times, so as to more accurately calculate the read noise model.

[0074] The higher the gain value of the camera is, the brighter the brightness of the first RAW image is; and the longer the exposure time is, the brighter the brightness of the first RAW image is.

[0075] like Figure 2 Shown is a first RAW image captured by the electronic device in a dark environment.

[0076] It is understandable that the RAW domain has four channels: R, GR, GB, and B. Figure 3 Shown is an image of one of the channels of the first RAW image.

[0077] Step 104 : outputting the read noise of each second RAW image by using a read noise model in a light environment; each second RAW image is acquired by a camera with different gain values.

[0078] The second RAW image is the original image data collected by the camera in a light environment. Similarly, the second RAW image is the original image data without any linear processing, so in the processing flow (pipeline) of the camera recording data, the bad pixel removal function can be turned on.

[0079] In a light environment, the camera is controlled to collect a plurality of second RAW images with different gain values, and each second RAW image is input into a read noise model, so that the read noise of each second RAW image can be output.

[0080] Step 106 : determining a shot noise model based on each second RAW image and the read noise corresponding to each second RAW image.

[0081] The shot noise model is a model of the shot noise of the output image.

[0082] Optionally, the electronic device can obtain pixel data of each second RAW image, determine the mean and variance based on the pixel data of each second RAW image, fit a curve between the mean and variance based on the mean and variance of the second RAW images and the corresponding read noise, and determine the slope of the curve; determine the shot noise model based on the slope and the gain value of the camera.

[0083] Step 108 : obtaining a target noise model of the RAW image captured by the camera based on the read noise model and the shot noise model.

[0084] The target noise model is a model for estimating the noise of the RAW image captured by the camera.

[0085] Optionally, the electronic device combines the read noise model and the shot noise model to obtain a target noise model.

[0086] In the above-mentioned noise model estimation method, the electronic device obtains multiple first RAW images captured by the camera at different gain values ​​in a dark environment, and determines a read noise model based on the multiple first RAW images. The objectively existing read noise generated by the camera in the dark environment can be obtained through the read noise model. In a light environment, the read noise of each second RAW image is first output through the read noise model, and then the shot noise model in the light environment can be determined based on each second RAW image and the read noise corresponding to each second RAW image. Then, based on the read noise model in the dark environment and the shot noise model in the light environment, the target noise model of the RAW image captured by the camera can be accurately estimated, and the noise of the RAW image captured by the camera can be more accurately estimated through the target noise model.

[0087] In one embodiment, a read noise model is determined based on multiple first RAW images, including: for each gain value corresponding to the first RAW image, obtaining a noise distribution parameter of the first RAW image corresponding to the gain value; the noise distribution parameter is used to simulate the read noise of the first RAW image; based on the correspondence between the gain value and the noise distribution parameter, generating a read noise model.

[0088] The noise distribution parameter is a parameter that characterizes the distribution of noise. The noise distribution parameter may specifically include at least one of a scale parameter and a shape parameter. The shape parameter indicates the type of noise distribution, and the scale parameter is the scale of the noise distribution curve.

[0089] For each gain value corresponding to the first RAW image, the electronic device obtains the noise distribution parameter of the first RAW image corresponding to the gain value, and a gain value corresponding to the first RAW image can obtain a noise distribution parameter. Then, based on the corresponding relationship between each gain value and the corresponding noise distribution parameter, a relationship curve between the gain value and the corresponding noise distribution parameter can be fitted, and the relationship curve is used as a read noise model.

[0090] It can be understood that the electronic device inputs the RAW image into the read noise model, the read noise model obtains the gain value corresponding to the RAW image, determines the noise distribution parameters corresponding to the gain value according to the relationship curve, and can simulate the read noise of the RAW image through the noise distribution parameters and output the read noise.

[0091] In this embodiment, for each gain value corresponding to the first RAW image, the noise distribution parameters of the first RAW image corresponding to the gain value are obtained. The noise distribution parameters are used to simulate the read noise of the first RAW image. Based on the correspondence between the gain value and the noise distribution parameters, an accurate read noise model can be generated.

[0092] In one embodiment, for each gain value corresponding to the first RAW image, obtaining noise distribution parameters of the first RAW image corresponding to the gain value includes: for each gain value corresponding to the first RAW image, obtaining multiple groups of noise distribution parameters, and substituting the multiple groups of noise distribution parameters into a preset noise distribution model respectively to obtain multiple candidate noise distribution models; determining a target noise distribution model from the multiple candidate noise distribution models, and using the noise distribution parameters of the target noise distribution model as the noise distribution parameters of the first RAW image corresponding to the gain value.

[0093] The preset noise distribution model may be Tukey-Lambda. Tukey-Lambda is a distribution family, which has a collection of multiple similar noise distribution models. The electronic device inputs different noise distribution parameters into the distribution group to obtain different noise distribution models. However, the first RAW image collected in a dark environment is not stable, so it is necessary to use different noise distribution parameters to construct multiple candidate noise distribution models, so as to determine the target noise distribution model.

[0094] Optionally, for each gain value corresponding to the first RAW image, multiple groups of noise distribution parameters within a preset selection range are traversed, and each group of noise distribution parameters is substituted into a preset noise distribution model to obtain a candidate noise distribution model; each candidate noise distribution model is compared with the first RAW image, and a target noise distribution model is determined based on the comparison results.

[0095] The electronic device determines the difference between each candidate noise distribution model and the first RAW image, and determines the target noise distribution model based on each difference value. Specifically, for each candidate noise distribution model, the electronic device compares the grayscale value of each pixel point preset in the candidate noise distribution model with the grayscale value of the corresponding pixel point in the first RAW image, determines the difference value of each corresponding pixel point, and adds the difference values ​​of each pixel point to obtain the difference value between the candidate noise distribution model and the first RAW image.

[0096] In one embodiment, the electronic device may use the candidate noise distribution model with the smallest difference value as the target noise distribution model. In another embodiment, the electronic device may use the candidate noise distribution model with the second smallest difference value as the target noise distribution model. In other embodiments, the electronic device may also determine the target noise distribution model in other ways, which are not limited here.

[0097] Optionally, the noise distribution parameter includes at least one of a scale parameter and a shape parameter, and the shape parameters corresponding to different gain values ​​are relatively stable, so the scale parameter in the noise distribution parameter of the first RAW image corresponding to the gain value can be obtained.

[0098] In this embodiment, for each gain value corresponding to the first RAW image, multiple groups of noise distribution parameters are obtained, and the multiple groups of noise distribution parameters are respectively substituted into the preset noise distribution model to obtain multiple candidate noise distribution models. Then, the electronic device can determine an accurate target noise distribution model from the multiple candidate noise distribution models, and use the noise distribution parameters of the target noise distribution model as the noise distribution parameters of the first RAW image corresponding to the gain value, thereby improving the accuracy of the generated read noise model.

[0099] In one embodiment, for each gain value corresponding to the first RAW image, obtaining the noise distribution parameter of the first RAW image corresponding to the gain value includes: for each gain value corresponding to the first RAW image, obtaining the noise distribution parameter of each channel in the first RAW image corresponding to the gain value; based on the correspondence between the gain value and the noise distribution parameter, generating a read noise model, including: based on the correspondence between the gain value and the noise distribution parameter of each channel, generating a first channel noise model for each channel; based on each first channel noise model, obtaining a read noise model.

[0100] It is understandable that the RAW domain of the image includes R, GR, GB and B channels, and there are some differences in the photosensitivity, process manufacturing and other aspects of the photosensitive elements in each channel, so different channels have inconsistent photon induction. Therefore, the electronic device obtains the noise distribution parameters of each channel in the first RAW image corresponding to the gain value, so as to more accurately determine the noise model of each channel.

[0101] The first channel noise model is a model for the read noise of one channel of the output image in the RAW domain.

[0102] The RAW domain includes R, GR, GB and B channels, and the first channel noise models of the four channels can be generated accordingly, namely the first channel noise model of the R channel, the first channel noise model of the GR channel, the first channel noise model of the GB channel and the first channel noise model of the B channel.

[0103] Optionally, for each channel, the electronic device may fit a relationship curve between the gain value and the noise distribution parameter of the channel based on the correspondence between each gain value and the noise distribution parameter of the channel, and use the relationship curve as the first channel noise model of the channel.

[0104] It can be understood that the electronic device inputs the RAW image into the first channel noise model of the specified channel, and the first channel noise model of the specified channel obtains the gain value corresponding to the RAW image, and determines the noise distribution parameters of the specified channel corresponding to the gain value based on the relationship curve between the gain value and the noise distribution parameters of the specified channel. The read noise of the specified channel of the RAW image can be simulated through the noise distribution parameters, and the read noise of the specified channel can be output.

[0105] Figure 4 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a scale parameter of an R channel in an embodiment. Figure 5 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a scale parameter of a GR channel in an embodiment. Figure 6 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a scale parameter of a GB channel in an embodiment. Figure 7 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a scale parameter of a B channel in an embodiment. Figure 4 , Figure 5 , Figure 6 and Figure 7 In the figure, the horizontal axis is the scale parameter and the vertical axis is the gain value.

[0106] Optionally, the electronic device may combine the first channel noise models to obtain a read noise model.

[0107] In this embodiment, the electronic device obtains the noise distribution parameters of each channel in the first RAW image corresponding to the gain value, and thus more accurately generates the first channel noise model of each channel based on the correspondence between the gain value and the noise distribution parameters of each channel. A more accurate read noise model can be obtained based on each first channel noise model, and thus a more accurate target noise model can be obtained, thereby avoiding the problem of low accuracy of the generated target noise model due to differences in the sensitivity, process manufacturing, etc. of the photosensitive elements in different channels.

[0108] In one embodiment, for each gain value corresponding to the first RAW image, obtaining the noise distribution parameters of each channel in the first RAW image corresponding to the gain value includes: for each gain value corresponding to the first RAW image, obtaining multiple groups of noise distribution parameters, and substituting the multiple groups of noise distribution parameters into a preset noise distribution model respectively to obtain multiple candidate noise distribution models; determining a target noise distribution model for each channel from the multiple candidate noise distribution models, and using the noise distribution parameters of the target noise distribution model of each channel as the noise distribution parameters of each channel of the first RAW image corresponding to the gain value.

[0109] In one embodiment, a shot noise model is determined based on each second RAW image and the read noise corresponding to each second RAW image, including: for a gain value corresponding to each second RAW image, based on each second RAW image and the read noise corresponding to each second RAW image, obtaining a shot noise parameter corresponding to the gain value; and generating a shot noise model based on a correspondence between the gain value and the shot noise parameter.

[0110] The second RAW image is raw image data collected by the camera in a light environment. The shot noise parameter is a parameter of a shot noise model.

[0111] Optionally, for each gain value corresponding to the second RAW image, a shot noise parameter corresponding to the gain value can be obtained; then, based on multiple groups of data, each group of data includes a gain value and a corresponding shot noise parameter, a relationship curve between the gain value and the shot noise parameter can be fitted, and the relationship curve is used as a shot noise model; wherein, the relationship curve can characterize the corresponding relationship between the gain value and the shot noise parameter.

[0112] In this embodiment, for the gain value corresponding to each second RAW image, the electronic device obtains the shot noise parameter corresponding to the gain value based on each second RAW image and the read noise corresponding to each second RAW image, so that the shot noise model can be accurately generated based on the correspondence between the gain value and the shot noise parameter.

[0113] In one embodiment, the second RAW image includes a plurality of color blocks; based on each second RAW image and the read noise corresponding to each second RAW image, a shot noise parameter corresponding to the gain value is obtained, including: dividing each color block into a plurality of small blocks, and obtaining a data set corresponding to each color block; each group of data sets includes a plurality of data pairs, and each data pair includes a mean and a variance of the first RAW data of each pixel point in each small block; obtaining the read noise corresponding to each second RAW image; for a group of data sets, using the mean of each small block in the data set as the abscissa, the variance as the ordinate, and the read noise as the intercept, to fit a candidate curve of the mean and the variance; based on each candidate curve, determining a target curve, and using the slope of the target curve as the shot noise parameter corresponding to the gain value.

[0114] In a light environment, the electronic device captures a standard image card to obtain a second RAW image; wherein the standard image card includes a plurality of color blocks of different colors. Figure 8 The figure shows a schematic diagram of a standard picture card in one embodiment. The standard picture card includes 24 color blocks of different colors. Then, the second RAW image obtained by the electronic device photographing the standard picture card also includes 24 color blocks of different colors.

[0115] Optionally, for the gain value corresponding to each second RAW image, the electronic device divides each color block in the second RAW image into multiple small blocks, and calculates the mean and variance of the first RAW data of each pixel in each small block respectively. Each small block corresponds to a pair of data pairs, and the data set corresponding to each color block can be obtained.

[0116] For a set of data sets {M, V}, with the mean of each small block in the data set as the abscissa M, the variance V as the ordinate, and the read noise base of the second RAW image as the intercept, a candidate curve of the mean and variance can be fitted.

[0117] It can be understood that the second RAW image includes multiple color blocks, each color block corresponds to a set of data sets, and each set of data sets can be fitted with a candidate curve of mean and variance, so the second RAW image can obtain multiple candidate curves.

[0118] In one embodiment, the electronic device may determine the candidate curve with the smallest slope from among the candidate curves as the target curve. In another embodiment, the electronic device may determine the smoothest candidate curve from among the candidate curves as the target curve. In other embodiments, the electronic device may also determine the target curve in other ways, which are not limited here.

[0119] In this embodiment, the electronic device divides each color block into multiple small blocks, and obtains a data set corresponding to each color block; each set of data sets includes multiple data pairs, and each data pair includes the mean and variance of the first RAW data of each pixel point in each small block; the read noise corresponding to each second RAW image is obtained; for a set of data sets, the mean of each small block in the data set is used as the horizontal coordinate, the variance is used as the vertical coordinate, and the read noise is used as the intercept to fit a candidate curve of the mean and variance; thereby, a more accurate target curve can be determined from each candidate curve, and the slope of the target curve is used as the shot noise parameter corresponding to the gain value, thereby improving the accuracy of the shot noise parameter.

[0120] In one embodiment, the method for acquiring the first RAW data of each pixel in each small block includes: acquiring the second RAW data of each pixel in each small block; and performing black level noise reduction processing on each second RAW data to obtain the first RAW data.

[0121] The second RAW data is original RAW data in the second RAW image. The first RAW data is RAW data that has been processed by black level noise reduction.

[0122] Defines the signal level corresponding to the image data being 0. Adjusting the black level does not affect the signal amplification factor, but only shifts the signal up and down. If the black level is adjusted upward, the image will become darker, and if the black level is adjusted downward, the image will become brighter.

[0123] The black level (also called absolute black level) setting is the lowest point of black. The lowest point of black is when the energy of the electron beam emitted from the CRT (Cathode Ray Tube) is lower than the basic energy that allows the phosphorescent material (fluorescent material) to start emitting light. The lowest black is displayed on the screen.

[0124] Black level, also known as dark level, is the dark level noise that objectively exists in the camera system.

[0125] Optionally, the electronic device subtracts a noise value of a black level from each second RAW data to obtain the first RAW data.

[0126] It can be understood that the electronic device can reduce the noise of each second RAW data by subjecting each second RAW data to black level noise reduction processing, obtain more accurate first RAW data, avoid the problem of large errors caused by using RAW data that has not been processed with black level noise reduction for calculation, and more accurately calculate the mean and variance of the first RAW data of each pixel in each small block.

[0127] In one embodiment, the target noise model may be: noiseLevel=k*(signal-bl)+base, where noiseLevel is the target noise output by the target noise model, k is the shot noise parameter, signal represents the pixel value of the RAW image, bl represents the noise value of the black level, k*(signal-bl) is the shot noise of the RAW image, and base is the read noise of the RAW image.

[0128] In one embodiment, for the gain value corresponding to each second RAW image, based on each second RAW image and the read noise corresponding to each second RAW image, a shot noise parameter corresponding to the gain value is obtained, including: for the gain value corresponding to each second RAW image, based on the RAW data of each channel of each second RAW image and the read noise corresponding to each second RAW image in each channel, the shot noise parameter of each channel corresponding to the gain value is obtained; based on the correspondence between the gain value and the shot noise parameter, a shot noise model is generated, including: based on the correspondence between the gain value and the shot noise parameter of each channel, a second channel noise model of each channel is generated; based on the second channel noise model, a shot noise model is obtained.

[0129] It is understandable that the RAW domain of the image includes R, GR, GB and B channels, and there are some differences in the photosensitivity and process manufacturing of the photosensitive elements in each channel, so different channels have inconsistent photon sensing. Therefore, the electronic device obtains the RAW data of each channel in the second RAW image corresponding to the gain value and the read noise of each channel, so as to more accurately obtain the shot noise parameters of each channel corresponding to the gain value.

[0130] The second channel noise model is a model for the shot noise of one channel of the output image in the RAW domain.

[0131] The RAW domain includes R, GR, GB and B channels, and the second channel noise models of the four channels can be generated accordingly, namely, the second channel noise model of the R channel, the second channel noise model of the GR channel, the second channel noise model of the GB channel and the second channel noise model of the B channel.

[0132] Optionally, for each channel, the electronic device may fit a relationship curve between the gain value and the shot noise parameter of each channel based on the corresponding relationship between the gain value and the shot noise parameter of the channel, and use the relationship curve as the second channel noise model of the channel.

[0133] Fig. 9 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of an R channel in an embodiment. Fig.10FIG. 1 is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a GR channel in an embodiment. Fig.11 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a GB channel in an embodiment. Fig.12 FIG. 1 is a schematic diagram of a relationship curve between a gain value and a shot noise parameter of a B channel in an embodiment. Fig. 9 , Fig.10 , Fig.11 and Fig.12 In the figure, the horizontal axis is the shot noise parameter, and the vertical axis is the gain value.

[0134] Optionally, the electronic device may combine the first channel noise models to obtain a read noise model.

[0135] In this embodiment, the electronic device obtains the shot noise parameters of each channel corresponding to the gain value, and thus more accurately generates the second channel noise model of each channel based on the correspondence between the gain value and the shot noise parameters of each channel. A more accurate shot noise model can be obtained based on each second channel noise model, and thus a more accurate target noise model can be obtained, thereby avoiding the problem of low accuracy of the generated target noise model due to differences in the sensitivity, process manufacturing, etc. of the photosensitive elements in different channels.

[0136] In one embodiment, based on the RAW data of each channel of each second RAW image and the read noise corresponding to each channel of each second RAW image, a shot noise parameter of each channel corresponding to a gain value is obtained, including: dividing each color block into a plurality of small blocks, and for each channel of the second RAW image, obtaining a data set corresponding to each color block; each set of data sets includes a plurality of data pairs, and each data pair includes a mean and a variance of the first RAW data of each pixel point in each small block; for each image, obtaining a read noise corresponding to each channel; in each channel, for a set of data sets, using the mean of each small block in the data set as the abscissa, the variance as the ordinate, and the read noise as the intercept, to fit a candidate curve of the mean and the variance; based on each candidate curve of each channel, determining a target curve of each channel, and using the slope of the target curve of each channel as the shot noise parameter of each channel corresponding to the gain value.

[0137] In one embodiment, Fig.13 As shown, an image processing method is provided. This embodiment uses the method applied to an electronic device as an example. The electronic device can be a terminal or a server. The above method can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server.

[0138] In this embodiment, the image processing method includes the following steps:

[0139] Step 1302: Acquire the original RAW image captured by the camera.

[0140] Step 1304, determining the target noise of the original RAW image through the target noise model, and processing the original RAW image based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0141] The electronic device inputs the original RAW image into a target noise model, through which the target noise model may be the target noise of the original RAW image, and processes the original RAW image based on the target noise to obtain the target RAW image, wherein the target noise may include read noise and shot noise.

[0142] The electronic device processes the original RAW image, including at least one of the following processing operations: noise reduction processing, sharpening processing, and de-mosaicing (AI-demosaic) processing.

[0143] It should be noted that the electronic device may also perform other processing on the original RAW image based on the target noise, which is not limited here.

[0144] like Fig.14 FIG. 1 is a schematic diagram of a target RAW image in an embodiment, wherein the target RAW image is obtained by processing the original RAW image based on the target noise, the target noise is determined by a target noise model, and the target noise model is obtained according to the above-mentioned noise model estimation method.

[0145] Optionally, the electronic device may train the target noise model to obtain a trained target noise model, and determine the target noise of the original RAW image through the trained target noise model.

[0146] In the above-mentioned image processing method, the electronic device obtains the original RAW image captured by the camera; the target noise model obtained by the above-mentioned noise model estimation method can more accurately determine the target noise of the original RAW image, so that the original RAW image can be more accurately processed based on the more accurate target noise to obtain a clearer target RAW image.

[0147] Moreover, based on more accurate target noise, electronic devices can enhance the resolution of images when processing original RAW images, effectively improve the extraction of image details, and simultaneously achieve image noise reduction and detail enhancement to obtain target RAW images with clearer details.

[0148] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0149] Based on the same inventive concept, the embodiment of the present application also provides a device for estimating a noise model for implementing the method for estimating a noise model involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of the device for estimating one or more noise models provided below can refer to the limitations of the method for estimating a noise model above, and will not be repeated here.

[0150] In one embodiment, Fig.15 As shown, a noise model estimation device is provided, including: a read noise model determination module 1502, a shot noise model determination module 1504 and a target noise model determination module 1506, wherein:

[0151] The read noise model determination module 1502 is used to obtain a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and determine a read noise model according to the plurality of first RAW images.

[0152] The shot noise model determination module 1504 is used to output the read noise of each second RAW image through the read noise model in a light environment; each second RAW image is acquired by the camera with different gain values; based on each second RAW image and the read noise corresponding to each second RAW image, the shot noise model is determined.

[0153] The target noise model determination module 1506 is used to obtain a target noise model of a RAW image acquired by the camera based on the read noise model and the shot noise model.

[0154] In the above-mentioned noise model estimation device, the electronic device obtains multiple first RAW images captured by the camera with different gain values ​​in a dark environment, and determines a read noise model based on the multiple first RAW images. The objectively existing read noise generated by the camera in the dark environment can be obtained through the read noise model. In a light environment, the read noise of each second RAW image is first output through the read noise model, and then the shot noise model in the light environment can be determined based on each second RAW image and the read noise corresponding to each second RAW image. Then, based on the read noise model in the dark environment and the shot noise model in the light environment, the target noise model of the RAW image captured by the camera can be accurately estimated, and the noise of the RAW image captured by the camera can be more accurately estimated through the target noise model.

[0155] In one embodiment, the read noise model determination module 1502 is further used to obtain the noise distribution parameters of the first RAW image corresponding to each gain value corresponding to the first RAW image; the noise distribution parameters are used to simulate the read noise of the first RAW image; and based on the correspondence between the gain value and the noise distribution parameter, a read noise model is generated.

[0156] In one embodiment, the above-mentioned read noise model determination module 1502 is also used to obtain multiple groups of noise distribution parameters for each gain value corresponding to the first RAW image, and substitute the multiple groups of noise distribution parameters into the preset noise distribution model respectively to obtain multiple candidate noise distribution models; determine the target noise distribution model from the multiple candidate noise distribution models, and use the noise distribution parameters of the target noise distribution model as the noise distribution parameters of the first RAW image corresponding to the gain value.

[0157] In one embodiment, the read noise model determination module 1502 is further used to obtain the noise distribution parameters of each channel in the first RAW image corresponding to the gain value of each first RAW image; based on the correspondence between the gain value and the noise distribution parameter of each channel, generate the first channel noise model of each channel; and based on each first channel noise model, obtain the read noise model.

[0158] In one embodiment, the shot noise model determination module 1504 is further used to obtain, for each gain value corresponding to the second RAW image, a shot noise parameter corresponding to the gain value based on each second RAW image and the read noise corresponding to each second RAW image; and generate a shot noise model based on the correspondence between the gain value and the shot noise parameter.

[0159] In one embodiment, the second RAW image includes multiple color blocks; the above-mentioned shot noise model determination module 1504 is also used to divide each color block into multiple small blocks, and obtain a data set corresponding to each color block; each group of data sets includes multiple data pairs, and each data pair includes the mean and variance of the first RAW data of each pixel point in each small block; obtain the read noise corresponding to each second RAW image; for a group of data sets, use the mean of each small block in the data set as the horizontal coordinate, the variance as the vertical coordinate, and the read noise as the intercept to fit a candidate curve of the mean and variance; based on each candidate curve, determine the target curve, and use the slope of the target curve as the shot noise parameter corresponding to the gain value.

[0160] In one embodiment, the shot noise model determination module 1504 is further used to obtain second RAW data of each pixel in each small block; and perform black level noise reduction processing on each second RAW data to obtain first RAW data.

[0161] In one embodiment, the shot noise model determination module 1504 is further used to obtain, for the gain value corresponding to each second RAW image, a shot noise parameter of each channel corresponding to the gain value based on the RAW data of each channel of each second RAW image and the read noise corresponding to each second RAW image in each channel; generate a second channel noise model for each channel based on the correspondence between the gain value and the shot noise parameter of each channel; and obtain a shot noise model based on each second channel noise model.

[0162] Based on the same inventive concept, the embodiment of the present application also provides an estimation device for a noise model for implementing the above-mentioned image processing method. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above-mentioned method, so the specific limitations in the one or more image processing device embodiments provided below can refer to the limitations on the image processing method above, and will not be repeated here.

[0163] In one embodiment, Fig.16 As shown, an image processing device is provided, including: an acquisition module 1602 and a processing module 1604, wherein:

[0164] The acquisition module 1602 is used to acquire the original RAW image captured by the camera.

[0165] The processing module 1604 is used to determine the target noise of the original RAW image through the target noise model, and process the original RAW image based on the target noise to obtain the target RAW image; wherein the target noise model is obtained according to the above-mentioned noise model estimation method.

[0166] The above-mentioned image processing device, the electronic device obtains the original RAW image captured by the camera; the target noise model obtained by the above-mentioned noise model estimation method can more accurately determine the target noise of the original RAW image, so that the original RAW image can be more accurately processed based on the more accurate target noise to obtain a clearer target RAW image.

[0167] Each module in the noise model estimation device or image processing device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0168] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.17 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a noise model estimation method or an image processing method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0169] Those skilled in the art will understand that Fig.17 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0170] The embodiment of the present application also provides a computer-readable storage medium. One or more non-volatile computer-readable storage media containing computer-executable instructions, when the computer-executable instructions are executed by one or more processors, the processors execute the steps of the noise model estimation method or the steps of the image processing method.

[0171] An embodiment of the present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the noise model estimation method or the steps of the image processing method.

[0172] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0173] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0174] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0175] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A noise model estimation method, characterized in that: include: Acquire a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and for each gain value corresponding to the first RAW image, acquire a noise distribution parameter of the first RAW image corresponding to the gain value; The noise distribution parameters are used to simulate the read noise of the first RAW image; generating a read noise model based on the corresponding relationship between the gain value and the noise distribution parameter; In a light environment, outputting the read noise of each second RAW image through the read noise model; Each of the second RAW images is acquired by the camera with different gain values; For each gain value corresponding to the second RAW image, based on each of the second RAW images and the read noise corresponding to each of the second RAW images, obtaining a shot noise parameter corresponding to the gain value; generating a shot noise model based on the corresponding relationship between the gain value and the shot noise parameter; Based on the read noise model and the shot noise model, a target noise model of the RAW image collected by the camera is obtained; the target noise model is noiseLevel=k*(signal-bl)+base, wherein noiseLevel is the target noise output by the target noise model, k is a shot noise parameter, signal represents the pixel value of the RAW image, bl represents the noise value of the black level, k*(signal-bl) is the shot noise of the RAW image, and base is the read noise of the RAW image.

2. The method according to claim 1, characterized in that: The step of obtaining, for each gain value corresponding to the first RAW image, a noise distribution parameter of the first RAW image corresponding to the gain value, includes: For each gain value corresponding to the first RAW image, multiple groups of noise distribution parameters are obtained, and the multiple groups of noise distribution parameters are respectively substituted into a preset noise distribution model to obtain multiple candidate noise distribution models; A target noise distribution model is determined from a plurality of candidate noise distribution models, and noise distribution parameters of the target noise distribution model are used as noise distribution parameters of the first RAW image corresponding to the gain value.

3. The method according to claim 1, characterized in that The step of obtaining, for each gain value corresponding to the first RAW image, a noise distribution parameter of the first RAW image corresponding to the gain value, includes: For each gain value corresponding to the first RAW image, obtaining a noise distribution parameter of each channel in the first RAW image corresponding to the gain value; The generating a read noise model based on the correspondence between the gain value and the noise distribution parameter comprises: Generate a first channel noise model for each channel based on the corresponding relationship between the gain value and the noise distribution parameter of each channel; Based on each of the first channel noise models, a read noise model is obtained.

4. The method according to claim 1, characterized in that: The second RAW image includes a plurality of color blocks; and obtaining the shot noise parameter corresponding to the gain value based on each of the second RAW images and the read noise corresponding to each of the second RAW images includes: Divide each color block into multiple small blocks, and obtain a data set corresponding to each color block; each set of the data sets includes multiple data pairs, and each data pair includes the mean and variance of the first RAW data of each pixel point in each small block; Acquire the read noise corresponding to each of the second RAW images; For a set of data sets, a candidate curve of the mean and variance is fitted by taking the mean of each small block in the data set as the abscissa, the variance as the ordinate, and the read noise as the intercept; Based on each candidate curve, a target curve is determined, and the slope of the target curve is used as a shot noise parameter corresponding to the gain value.

5. The method according to claim 4, characterized in that The method for acquiring the first RAW data of each pixel in each small block includes: Obtain the second RAW data of each pixel in each small block; Each second RAW data is subjected to black level noise reduction processing to obtain first RAW data.

6. The method according to claim 4, characterized in that The step of obtaining, for each gain value corresponding to the second RAW image, a shot noise parameter corresponding to the gain value based on each of the second RAW images and a read noise corresponding to each of the second RAW images, includes: For each gain value corresponding to the second RAW image, based on the RAW data of each channel of each second RAW image and the read noise corresponding to each channel of each second RAW image, obtaining a shot noise parameter of each channel corresponding to the gain value; The generating a shot noise model based on the correspondence between the gain value and the shot noise parameter comprises: Based on the corresponding relationship between the gain value and the shot noise parameter of each channel, generating a second channel noise model for each channel; Based on each of the second channel noise models, a shot noise model is obtained.

7. An image processing method, characterized in that: include: Get the original RAW image captured by the camera; The target noise of the original RAW image is determined by a target noise model, and the original RAW image is processed based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to the method according to any one of claims 1 to 6.

8. A noise model estimation device, characterized in that: include: A read noise model determination module is used to obtain a plurality of first RAW images captured by a camera with different gain values ​​in a dark environment, and for each gain value corresponding to the first RAW image, obtain a noise distribution parameter of the first RAW image corresponding to the gain value; The noise distribution parameters are used to simulate the read noise of the first RAW image; generating a read noise model based on the corresponding relationship between the gain value and the noise distribution parameter; a shot noise model determination module, configured to output the read noise of each second RAW image through the read noise model in a light environment; each second RAW image is acquired by the camera with different gain values; for a gain value corresponding to each second RAW image, based on each second RAW image and the read noise corresponding to each second RAW image, obtain a shot noise parameter corresponding to the gain value; and generate a shot noise model based on a corresponding relationship between the gain value and the shot noise parameter; A target noise model determination module is used to obtain a target noise model of the RAW image collected by the camera based on the read noise model and the shot noise model; the target noise model is noiseLevel=k*(signal-bl)+base, wherein noiseLevel is the target noise output by the target noise model, k is a shot noise parameter, signal represents the pixel value of the RAW image, bl represents the noise value of the black level, k*(signal-bl) is the shot noise of the RAW image, and base is the read noise of the RAW image.

9. An image processing device, characterized in that: include: The acquisition module is used to obtain the original RAW image captured by the camera; A processing module, used to determine the target noise of the original RAW image through a target noise model, and process the original RAW image based on the target noise to obtain a target RAW image; wherein the target noise model is obtained according to any one of claims 1 to 6.

10. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor is caused to perform the steps of the noise model estimation method according to any one of claims 1 to 6.

11. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the computer program is executed by the processor, the processor is caused to perform the steps of the image processing method according to any one of claim 7.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 or 7 are implemented.

13. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 or 7 are implemented.

Images