Image noise evaluation method and device, electronic equipment and storage medium
By calculating the difference between the denoised image and the original image, the sum of the number of non-0 pixels and pixel values of the denoised image is evaluated, and the problems of high difficulty and low accuracy of noise intensity quantization are solved, and efficient and accurate noise evaluation is achieved.
Patent Information
- Application Number
- CN202510326318.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-07-04
AI Technical Summary
In the prior art, the quantification of noise intensity is difficult and low in accuracy, and depends on the developer's subjective evaluation, resulting in high labor costs and affecting the training effect of deep learning models.
By acquiring the original image and the corresponding denoising image, the difference between the denoising image and the original image is calculated, and the noise fraction is calculated to evaluate the noise intensity and avoid subjective annotation using the ratio of the number of non-0 pixels in the denoising image.
It realizes efficient and accurate noise evaluation, saves labor costs, accurately locates noise distribution, and improves the accuracy of noise evaluation.
Smart Images
Figure CN120259224A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of data processing, and particularly to an image noise evaluation method, apparatus, electronic device, and storage medium. Background Art
[0002] Noise evaluation is widely applied to the basic processing algorithms of major video websites. By accurately identifying the frames with high noise intensity in a video and denoising them, effects such as old film restoration, reduction of abnormal phenomena such as compression distortion or edge blurring, and improvement of video coding efficiency can be achieved.
[0003] In the prior art, in order to train a deep learning model to identify and process noise in a video, first, a training data set is required. The training data set consists of a series of images, and each image is manually labeled with its corresponding noise intensity. The performance of the deep learning model depends to a large extent on the quality and quantity of the training data set.
[0004] However, currently, the noise intensity is usually based on the subjective evaluation of developers. Different people have significant differences in the noise evaluation of the same image. Therefore, it is difficult to quantify the noise intensity standard with low precision and requires a large amount of manpower. Summary of the Invention
[0005] To solve the above technical problems, the present application discloses an image noise evaluation method, apparatus, electronic device, and storage medium to at least solve the problem that the noise evaluation method of manually labeling noise in related technologies depends on the subjective evaluation of developers, the quantification of the noise intensity standard is difficult and has low precision, and a large amount of manpower is required. The technical solutions of the present disclosure are as follows:
[0006] In a first aspect, the present application discloses an image noise evaluation method, and the method includes:
[0007] Obtain an original image and a denoised image corresponding to the original image, where the denoised image is obtained after the original image is denoised;
[0008] Compare the denoised image with the original image to obtain a first noise image, where the first noise image is used to indicate the noise in the original image;
[0009] Take the number of non - zero pixels in the first noise image as a first quantity, and take the sum of the pixel values of the pixel points in the first noise image as a second quantity. Based on the ratio of the second quantity to the first quantity, calculate the noise score of the original image, where the noise score is used to evaluate the noise intensity of the original image.
[0010] Optionally, the step of comparing the denoised image with the original image to obtain a first noise image includes:
[0011] Calculate the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain a second noise image;
[0012] Extract the texture information of the denoised image to obtain a texture image;
[0013] In the second noise image, set the pixel values at the corresponding positions of the non-zero pixels in the texture image to 0 to obtain a first noise image.
[0014] Optionally, the calculating the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain a second noise image includes:
[0015] Calculate the difference between the pixel values of the pixels at the same corresponding positions in the denoised image and the original image;
[0016] Determine the absolute value of the difference to obtain a second noise image.
[0017] Optionally, the setting the pixel values at the corresponding positions of the non-zero pixels in the texture image to 0 in the second noise image to obtain a first noise image includes:
[0018] Set the pixels with pixel values less than a first threshold in the texture image to 0 to obtain a reference image, and determine the target positions of the non-zero pixels in the reference image;
[0019] In the second noise image, set the pixel values at the target positions to 0 to obtain a first noise image.
[0020] Optionally, the obtaining the original image and the denoised image corresponding to the original image includes:
[0021] Obtain the original image;
[0022] Input the original image into a pre-trained denoising model to obtain a denoised image;
[0023] Wherein, the denoising model is trained by the following steps:
[0024] Obtain a filtered image, and add noise to the filtered image to obtain an input image, where the filtered image is obtained by filtering a training image;
[0025] Input the input image into a preset deep learning model for denoising processing to obtain an output image;
[0026] Calculate the loss value between the output image and the filtered image. When the loss value does not meet the preset threshold, adjust the model parameters of the preset deep learning model until the loss value meets the preset threshold, and use the adjusted preset deep learning model as the denoising model.
[0027] Optionally, determining the noise score of the original image based on the ratio of the second quantity to the first quantity includes:
[0028] Divide the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter; the preset value is greater than 1;
[0029] When the value of the candidate parameter is greater than or equal to 1, set the noise score of the original image to 1;
[0030] When the value of the candidate parameter is less than 1, use the candidate parameter as the noise score of the original image.
[0031] Optionally, before dividing the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter, it further includes:
[0032] Judge whether the first quantity is greater than the product of the total number of pixel values in the first noise image and a preset ratio;
[0033] If it is less than the product, set the noise score of the original image to 0;
[0034] If it is not less than the product, perform the step of dividing the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter.
[0035] In a second aspect, an embodiment of the present invention provides an image noise evaluation device, including:
[0036] An acquisition module, configured to acquire an original image and a denoised image corresponding to the original image, where the denoised image is obtained by performing denoising processing on the original image;
[0037] A comparison module, configured to compare the denoised image with the original image to obtain a first noise image, where the first noise image is used to indicate the noise in the original image;
[0038] A calculation module, configured to use the number of non-0 pixels in the first noise image as the first quantity, the sum of the pixel values of the pixel points in the first noise image as the second quantity, and calculate the noise score of the original image based on the ratio of the second quantity to the first quantity, where the noise score is used to evaluate the noise intensity of the original image.
[0039] In a third aspect, an embodiment of the present invention provides an electronic device, including:
[0040] a processor;
[0041] a memory for storing executable instructions of the processor;
[0042] wherein, the processor is configured to execute the instructions to implement the image noise evaluation method described in any one of the above.
[0043] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an image noise evaluation electronic device, the image noise evaluation electronic device can execute the image noise evaluation method described in any one of the above.
[0044] Compared with the prior art, the present application has the following advantages:
[0045] In the present application, a raw image and a denoised image corresponding to the raw image are obtained. The denoised image is obtained after the raw image is denoised. The denoised image is compared with the raw image to obtain a first noise image, and the first noise image is used to indicate the noise in the raw image. The number of non-zero pixels in the first noise image is used as a first quantity, and the sum of the pixel values of the pixel points in the first noise image is used as a second quantity. Based on the ratio of the second quantity to the first quantity, the noise score of the raw image is calculated, and the noise score is used to evaluate the noise intensity of the raw image.
[0046] In this way, by comparing the raw image with the denoised image obtained after denoising the raw image, the noise existing in the raw image is identified to obtain the first noise image. Furthermore, by analyzing the number of non-zero pixels and the sum of the pixel values of each pixel point in the first noise image, the noise evaluation of the raw image can be realized, and the noise score of the raw image can be obtained. That is to say, the present application uses the denoising result of the raw image to detect the noise of the raw image in reverse, accurately locates the noise distribution of the raw image, and at the same time avoids subjective noise annotation, saves labor costs, and can realize efficient and accurate noise evaluation. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 is a flowchart of the steps of an image noise evaluation method of the present application;
[0048] Figure 2 is a logical schematic diagram of an image noise evaluation method of the present application;
[0049] Figure 3 is a structural block diagram of an image noise evaluation device of the present application;
[0050] Figure 4It is a schematic diagram of an electronic device of the present application;
[0051] Figure 5 It is a block diagram of a device for image noise evaluation of the present application. Specific embodiments
[0052] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0053] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.
[0054] Referring to Figure 1 , a step flowchart of an image noise evaluation method of the present application is shown, which may specifically include the following steps:
[0055] In step S11, an original image and a denoised image corresponding to the original image are obtained, and the denoised image is obtained after the original image is denoised.
[0056] In the related art, the noise evaluation of images is usually based on the subjective evaluation of developers. Different people have large differences in the noise evaluation of the same image. Therefore, it is difficult to quantify the noise intensity standard and the accuracy is low, and a large amount of manpower is required, which is not conducive to the training of deep learning models, resulting in the difficulty of the trained deep learning models to accurately evaluate the image noise.
[0057] Based on this, the present application provides an image noise evaluation method that does not rely on subjective annotation to solve the above problems.
[0058] First, in the present application, an original image and a denoised image corresponding to the original image need to be obtained. Specifically, the original image can be obtained through an image acquisition device (such as a camera, a scanner, etc.). The resolution and image format of the original image can be determined according to requirements, and the present application does not limit this. The denoised image is obtained by processing the original image through a certain algorithm, and its purpose is to remove the noise in the original image and improve the signal-to-noise ratio and visual effect of the original image.
[0059] Among them, there are many denoising algorithms, such as mean filtering, median filtering, Gaussian filtering, bilateral filtering, non-local mean filtering, etc. These denoising algorithms have their own advantages and disadvantages and are suitable for different types of noise and image characteristics. The appropriate denoising algorithm can be selected according to the noise situation of the original image and the requirements of subsequent tasks. For example, for an original image with rich texture and details, bilateral filtering can be selected; for an original image that requires high precision, non-local mean filtering can be selected.
[0060] In step S12, the denoised image is compared with the original image to obtain a first noise image, which is used to indicate the noise in the original image.
[0061] In this step, the first noise image can be obtained by comparing the denoised image with the original image. Since the denoised image is obtained by denoising the original image, the difference between the two lies in the noise existing in the original image, that is, the first noise image.
[0062] Among them, the process of comparing the denoised image with the original image can usually be achieved by calculating the difference in pixel values at each corresponding position between them. That is to say, the larger the pixel value of a certain pixel point in the first noise image, the stronger the noise at that position in the original image.
[0063] In step S13, the number of non-0 pixels in the first noise image is taken as the first quantity, and the sum of the pixel values of the pixel points in the first noise image is taken as the second quantity. Based on the ratio of the second quantity to the first quantity, the noise score of the original image is calculated, and the noise score is used to evaluate the noise intensity of the original image.
[0064] In this step, the number of non-0 pixels in the first noise image, that is, the number of noise pixels, can be taken as the first quantity. If the number of non-0 pixels is large, it means there is more noise in the original image. Conversely, if the number is small, it may indicate less noise in the original image. Therefore, the first quantity reflects the "quantity level" or "density" of the noise in the original image and can be denoted as noise_pix_num.
[0065] At the same time, the sum of the pixel values of the pixel points in the first noise image can also be calculated as the second quantity. The second quantity reflects the overall intensity of the noise in the image. The larger the second quantity, the stronger or more significant the noise in the original image; conversely, it may indicate weaker or more concealed noise in the original image. Therefore, the second quantity reflects the "intensity" or "total energy" of the noise in the original image and can be denoted as noise_sum.
[0066] Then, based on the ratio of the second quantity to the first quantity, the noise score of the original image can be determined. The higher the ratio of the second quantity to the first quantity, the stronger or denser the noise in the original image. The noise score provides a comprehensive measure that takes into account both the quantity and intensity of the noise, and thus can reflect the overall noise situation in the original image.
[0067] As can be seen from the above, the technical solution provided by the embodiments of the present disclosure identifies the noise existing in the original image by comparing the original image and the denoised image obtained after denoising the original image, and obtains the first noise image. Furthermore, by analyzing the number of non-zero pixels in the first noise and the sum of the pixel values of each pixel point, the noise evaluation of the original image can be realized, and the noise score of the original image can be obtained. That is to say, this application uses the denoising result of the original image to reverse-detect the noise of the original image, accurately locates the noise distribution of the original image, while avoiding subjective noise annotation, saving labor costs, and can achieve efficient and accurate noise evaluation.
[0068] In one implementation, the step of obtaining the original image and the denoised image corresponding to the original image in step S11 includes:
[0069] Obtain the original image;
[0070] Input the original image into a pre-trained denoising model to obtain the denoised image.
[0071] That is to say, the original image can be input into a pre-trained denoising model to obtain the corresponding denoised image.
[0072] Among them, the denoising model can be trained by the following steps:
[0073] Obtain a filtered image, and add noise to the filtered image to obtain an input image, where the filtered image is obtained by filtering the training image;
[0074] Input the input image into a preset deep learning model for denoising processing to obtain an output image;
[0075] Calculate the loss value between the output image and the filtered image. In the case where the loss value does not meet the preset threshold, adjust the model parameters of the preset deep learning model until the loss value meets the preset threshold, and use the adjusted preset deep learning model as the denoising model.
[0076] That is to say, first, training images can be obtained to construct a training data set. After filtering these training images, filtered images with noise removed can be obtained. For example, non-local mean filtering or bilateral filtering can be used, and the specific method is not limited. Then, noise is added to the filtered images to obtain input images. Furthermore, a deep learning model can be preset, and the input images are input into the preset deep learning network, using the filtered images as supervision signals to train the denoising model.
[0077] Specifically, the loss value between the output image and the filtered image can be calculated. When the loss value does not meet the preset threshold, it indicates that there is a large difference between the output image and the filtered image, that is, the denoising effect of the preset deep learning model on the input image is not good. Then, the model parameters of the preset deep learning model can be adjusted until the loss value meets the preset threshold, indicating that the difference between the output image and the filtered image is small, that is, the denoising effect of the preset deep learning model on the input image has met the expectation. Then, the adjusted preset deep learning model can be used as the denoising model.
[0078] In the embodiments of the present application, the denoising model can be constructed using deep learning technologies, such as Convolutional Neural Networks (CNN) or Encode-Decode structures, etc., and trained with a large number of noisy and noiseless images to learn the mapping relationship from noisy images to noiseless images.
[0079] In this way, when the original image is input into the denoising model, the denoising model can regard the original image as a noisy image, and then process it to output a denoised image corresponding to the original image without noise.
[0080] In one implementation, the step of comparing the denoised image with the original image in step S12 to obtain the first noise image includes:
[0081] Calculating the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain a second noise image;
[0082] Extracting the texture information of the denoised image to obtain a texture image;
[0083] In the second noise image, setting the pixel values at the positions corresponding to the non-zero pixels in the texture image to 0 to obtain the first noise image.
[0084] That is to say, in the process of comparing the denoised image with the original image, first, the difference between each corresponding pixel value of the denoised image and the original image can be calculated to obtain a second noise image that preliminarily reflects the denoising effect. The second noise image shows the change in pixel values in the original image before and after the denoising process, and these changes largely reflect the removal of noise in the original image.
[0085] Then, useful texture information can be extracted from the denoised image. Texture information is an important feature in an image, which describes the pixel arrangement in a local area of the image. The extracted texture information is saved as a new image, namely the texture image. Among them, there are many methods for extracting texture information, such as using filters, statistical methods, or machine learning-based algorithms, etc. This application does not limit this.
[0086] Furthermore, by combining the information of the second noise image and the texture image, the denoised image is compared with the original image. Specifically, in the second noise image, find the pixel positions in the texture image that are not 0, and set the pixel values at these positions to 0 to obtain the first noise image, so as to remove the noise that is considered to be the texture part of the image in the second noise image, because these parts should be retained during the denoising process rather than being removed as noise.
[0087] In this way, the first noise image not only contains the change information of pixel values in the original image before and after the denoising process, but also removes the noise that is considered to be the texture part of the image by combining the information of the texture image, and can more accurately reflect the noise existing in the original image.
[0088] In one implementation, the step of calculating the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain the second noise image in the above implementation includes:
[0089] Calculate the difference between the pixel values of the pixels at the same corresponding positions in the denoised image and the original image;
[0090] Determine the absolute value of the obtained difference to get the second noise image.
[0091] That is to say, the absolute value of the difference between the pixels at each corresponding position in the denoised image and the original image can be calculated to obtain a matrix with the same size as the original image, and this matrix is the second noise image. If the pixel values in the second noise image are generally small, it indicates that the difference between the denoised image and the original image is small, that is, there is a large amount of noise in the original image; on the contrary, if the pixel values are large, it indicates that the difference between the denoised image and the original image is large, that is, there is less noise in the original image.
[0092] The second noise image can provide an intuitive way to quantify the noise in the original image, which can be used to analyze the distribution of the noise and provide reference information for subsequent processing.
[0093] In one implementation, in the second noise image, the pixel values at the corresponding positions of the non-zero pixels in the texture image are set to 0 to obtain the first noise image, including:
[0094] In the texture image, the pixels with pixel values less than the first threshold are set to 0 to obtain a reference image, and the target positions where the pixels in the reference image are non-zero are determined;
[0095] In the second noise image, the pixel values at the target positions are set to 0 to obtain the first noise image.
[0096] That is to say, after obtaining the texture image, each pixel in the texture image can be traversed. If the pixel value is less than a preset first threshold, the value of the pixel is set to 0, so as to obtain a reference image that only contains pixel values greater than or equal to the first threshold.
[0097] Among them, the first threshold is set to a threshold that can distinguish texture and noise, usually an empirical value set manually. If the first threshold is set too low, some useful texture information may be misidentified as noise and removed; if the first threshold is set too high, some unnecessary noise may be retained. For example, the first threshold can be set to 30.
[0098] Then, all the position coordinates (x, y) where the pixel values in the reference image are non-zero can be recorded, and each pixel in the second noise image is traversed. The pixel value at the position (x, y) on the second noise image is set to 0 to obtain the first noise image, and the noise in the first noise image affects the visual perception of the picture.
[0099] In this way, the information of the texture image can be used to guide the modification of the second noise image. By removing the noise that is considered to be the texture part, such as the texture of cloth, wood grain, water ripple, etc., the noise that affects the visual perception of the original image can be seen more clearly, and a first noise image that more accurately reflects the noise existing in the original image can be obtained.
[0100] In one implementation, the step of determining the noise score of the original image based on the ratio of the second quantity to the first quantity in step S13 includes:
[0101] Dividing the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter; the preset value is greater than 1;
[0102] When the value of the candidate parameter is greater than or equal to 1, the noise score of the original image is set to 1;
[0103] When the value of the candidate parameter is less than 1, the candidate parameter is used as the noise score of the original image.
[0104] That is to say, the candidate parameter f of the original image can be determined by the following formula:
[0105] f = noise_sum / (a × noise_pix_num)
[0106] Where a represents a preset value. By setting the preset value, the range of the noise score can be adjusted, reducing a potentially very high ratio to a more manageable and judgmental range, making it more in line with subsequent judgment criteria and helping to more accurately judge the value of the noise score in subsequent steps.
[0107] Then, the following judgment can be made based on the value of the candidate parameter:
[0108] If the value of the candidate parameter is greater than or equal to 1, the noise score of the original image is set to 1, which means that the noise intensity in the original image is so high that it cannot be represented by a more precise noise score.
[0109] If the value of the candidate parameter is less than 1, the candidate parameter itself is used as the noise score of the original image, indicating that the noise intensity in the original image is below the preset threshold and the noise score can more precisely reflect the actual intensity of the noise.
[0110] In this way, by calculating the ratio of the second quantity to the first quantity and adjusting it through the preset value, a noise score that can reflect the overall noise situation in the image is finally obtained. The noise score not only considers the quantity of the noise but also the intensity of the noise, so it can more comprehensively evaluate the noise situation of the original image.
[0111] In one implementation, before the step of dividing the ratio of the second quantity to the first quantity by the preset value to obtain the candidate parameter in the above implementation, it further includes:
[0112] Judging whether the first quantity is greater than the product of the total number of pixel values in the first noise image and the preset ratio;
[0113] If it is less than the product, the noise score of the original image is set to 0;
[0114] If it is not less than the product, execute the step of dividing the ratio of the second quantity to the first quantity by the preset value to obtain the candidate parameter.
[0115] That is to say, before calculating the candidate parameter, it is also possible to first determine whether the first quantity is greater than the product of the total number of pixel values in the first noise image and the preset ratio. If the first quantity is less than this product, it indicates that the amount of noise in the original image is relatively small. In this case, the noise score of the original image can be set to 0, indicating that the noise in the original image is very low or almost non-existent.
[0116] If the first quantity is not less than this product, it indicates that the amount of noise in the original image is relatively large. In this case, the subsequent steps can be continued, that is, calculate the ratio of the second quantity to the first quantity and divide it by the preset value to obtain the candidate parameter, and then determine the noise score of the original image according to the value of the candidate parameter.
[0117] Among them, the preset ratio is a key parameter, which can be determined according to the actual application scenario and the characteristics of the image. For example, if you want to control the noise score more strictly, you can choose a smaller preset ratio; if you want to control the noise score more loosely, you can choose a larger preset ratio.
[0118] In this way, before calculating the candidate parameter, a preliminary evaluation of the amount of noise in the original image is first performed. If the amount of noise is very small, there is no need to perform subsequent calculations, and the noise score can be directly set to 0, thereby saving computing resources and improving processing efficiency.
[0119] For example, if the preset value is taken as 3 and the preset ratio is taken as 50%, then first, the number of pixels with non-zero pixel values in the first noise image, that is, the number of noise pixels, can be counted and denoted as noise_pix_num. Then, the sum of the pixel values of all pixels in the first noise image is calculated and denoted as noise_sum.
[0120] Then, it is determined whether the value of noise_pix_num is lower than the width × height × 50% of the first noise image. If the value of noise_pix_num is lower than this product, that is, when the number of noise pixels in the first noise image does not exceed 50%, it means that more than 50% of the pixels in the first noise image are non-noise pixels such as textures that do not need to be changed. At this time, it is considered that the noise has limited impact on the subjective perception, and it is determined that the picture does not contain noise, and the noise score can be taken as 0.
[0121] If the value of noise_pix_num is not lower than this product, it means that more than 50% of the pixels in the first noise image are noise pixels that need to be changed. At this time, it is considered that the noise has a greater impact on the subjective perception, and then the candidate parameter f = noise_sum / (3 × noise_pix_num) is calculated.
[0122] When the calculated result of f is greater than or equal to 1, the value of the noise score is set to 1. That is to say, when the average noise value per pixel in the original image, which is the ratio of the sum of the pixel values of all pixels in the first noise image to the number of noise pixels, reaches 3 or more, it is considered that there is serious noise in the picture, and the noise score takes the value of 1.
[0123] When the calculated result of f is less than 1, the value range of the noise score is between 0 and 1. The larger the value of the noise score, the greater the noise in the original image.
[0124] As Figure 2 shown, it is a schematic diagram of a specific business implementation example of the image noise evaluation method in this application, which includes the following steps:
[0125] Input the original image A into the pre-trained denoising model, and the output result is the denoised image B;
[0126] Compare the denoised image B and the original image A, calculate the difference between the pixel values of the corresponding same-position pixels in the denoised image B and the original image A, and determine the absolute value of the difference to obtain the second noise image C;
[0127] And, use the edge detection algorithm based on the Sobel operator to extract the edge texture on the denoised image B to obtain the texture image D;
[0128] In the texture image D, change the texture with a pixel value less than 30 to 0 to reduce noise interference, record all positions where the pixel value of the texture image D is not 0, and on the second noise image C, set the pixel values of these positions to 0 to obtain the first noise image E;
[0129] Furthermore, the noise of the original image can be quantified based on the first noise image E. Take the number of non-0 pixels in the first noise image E as the first quantity, and the sum of the pixel values of the pixel points in the first noise image E as the second quantity;
[0130] Judge whether the first quantity is greater than the product of the total number of pixel values in the first noise image E and the preset ratio. For example, the preset ratio can be 50%. Then, if the first quantity is less than 50% of the total number of pixel values in the first noise image E, set the noise score of the original image to 0; if the first quantity is not less than 50% of the total number of pixel values in the first noise image E, then divide the ratio of the second quantity to the first quantity by the preset value to obtain the candidate parameter;
[0131] For example, if the preset value can be 3, then divide the ratio of the second quantity to the first quantity by 3 to obtain a candidate parameter. When the value of the candidate parameter is greater than or equal to 1, set the noise score of the original image A to 1. When the value of the candidate parameter is less than 1, use the candidate parameter as the noise score of the original image A.
[0132] As can be seen from the above, the technical solution provided by the embodiments of the present disclosure identifies the noise existing in the original image by comparing the original image with the denoised image obtained after denoising the original image, and obtains the first noise image. Furthermore, by analyzing the number of non-0 pixels and the sum of the pixel values of each pixel point in the first noise, the noise evaluation of the original image can be realized, and the noise score of the original image can be obtained. That is to say, this application uses the denoising result of the original image to reversely detect the noise of the original image, accurately locates the noise distribution of the original image, avoids subjective noise annotation at the same time, saves labor costs, and can realize efficient and accurate noise evaluation.
[0133] Referring to Figure 3 , a schematic structural diagram of an image noise evaluation device of the present application is shown, which may specifically include:
[0134] An acquisition module 201, configured to acquire an original image and a denoised image corresponding to the original image, where the denoised image is obtained after the original image is denoised;
[0135] A comparison module 202, configured to compare the denoised image with the original image to obtain a first noise image, where the first noise image is used to indicate the noise in the original image;
[0136] A calculation module 203, configured to use the number of non-0 pixels in the first noise image as a first quantity, and the sum of the pixel values of the pixel points in the first noise image as a second quantity, and calculate the noise score of the original image based on the ratio of the second quantity to the first quantity, where the noise score is used to evaluate the noise intensity of the original image.
[0137] As can be seen from the above, the technical solution provided by the embodiments of the present disclosure identifies the noise existing in the original image by comparing the original image with the denoised image obtained after denoising the original image, and obtains the first noise image. Furthermore, by analyzing the number of non-0 pixels and the sum of the pixel values of each pixel point in the first noise, the noise evaluation of the original image can be realized, and the noise score of the original image can be obtained. That is to say, this application uses the denoising result of the original image to reversely detect the noise of the original image, accurately locates the noise distribution of the original image, avoids subjective noise annotation at the same time, saves labor costs, and can realize efficient and accurate noise evaluation.
[0138] For the device embodiments, since they are basically similar to the method embodiments, they are described relatively simply. For the relevant parts, refer to the corresponding descriptions in the method embodiments.
[0139] Figure 4 FIG. is a block diagram of an electronic device for image noise evaluation according to an exemplary embodiment, including a processor and a memory. The memory is used to store a computer program, and the processor is used to execute the program stored on the memory.
[0140] The memory may include a random access memory (RAM) and may also include a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.
[0141] The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
[0142] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided. For example, a memory including instructions, and the above instructions can be executed by the processor of the electronic device to complete the above method. Optionally, the computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.
[0143] In an exemplary embodiment, a computer program product is also provided. When it runs on a computer, it enables the computer to implement the above method for evaluating image noise.
[0144] As can be seen from the above, the technical solution provided by the embodiment of the present invention identifies the noise existing in the original image by comparing the original image and the denoised image obtained after denoising the original image, and obtains the first noise image. Furthermore, by analyzing the number of non-zero pixels and the sum of the pixel values of each pixel point in the first noise, the noise evaluation of the original image can be realized, and the noise score of the original image can be obtained. That is to say, this application uses the denoising result of the original image to reverse-detect the noise of the original image, accurately locates the noise distribution of the original image, and at the same time avoids subjective noise annotation, saves labor costs, and can realize efficient and accurate noise evaluation.
[0145] Figure 5 FIG. 4 is a block diagram of an apparatus 800 for image noise evaluation according to an exemplary embodiment.
[0146] For example, the apparatus 800 can be a mobile phone, a computer, a digital broadcast electronic device, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.
[0147] Referring to Figure 5 , the apparatus 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.
[0148] The processing component 802 generally controls the overall operation of the apparatus 800, such as operations associated with display, telephone calls, data communication, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above methods. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.
[0149] The memory 804 is configured to store various types of data to support the operation of the device 800. Examples of these data include instructions for any application or method operating on the apparatus 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk.
[0150] The power supply component 807 provides power for various components of the device 800. The power supply component 807 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 800.
[0151] The multimedia component 808 includes a screen that provides an output interface between the device 800 and an account. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from an account. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of a touch or swipe action but also detect the duration and pressure associated with the touch or swipe operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data to be processed. Each of the front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capabilities.
[0152] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when the device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 further includes a speaker for outputting audio signals.
[0153] The I / O interface 812 provides an interface between the processing component 802 and a peripheral interface module, and the peripheral interface module may be a keyboard, a click wheel, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.
[0154] The sensor assembly 814 includes one or more sensors for providing a status assessment of various aspects of the device 800. For example, the sensor assembly 814 can detect the on / off state of the device 800, the relative positioning of components, such as the display and keypad of the device 800. The sensor assembly 814 can also detect a change in the position of the device 800 or a component of the device 800, the presence or absence of contact of an object with the device 800, the orientation or acceleration / deceleration of the device 800, and a change in the temperature of the device 800. The sensor assembly 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 can also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 can also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
[0155] The communication component 816 is configured to facilitate communication between the device 800 and other devices in a wired or wireless manner. The device 800 can access a wireless network based on communication standards, such as WiFi, a carrier network (such as 2G, 3G, 4G, or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
[0156] In an exemplary embodiment, the device 800 can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the methods described in the first aspect and the second aspect.
[0157] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions, such as the memory 804 including instructions, is also provided. The above instructions can be executed by the processor 820 of the device 800 to complete the above method. Optionally, for example, the storage medium can be a non-transitory computer-readable storage medium. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.
[0158] In an exemplary embodiment, there is also provided a computer program product including instructions that, when running on a computer, cause the computer to execute the image noise evaluation method described in any one of the above embodiments.
[0159] As can be seen from the above, the technical solution provided by the embodiments of the present invention identifies the noise existing in the original image by comparing the original image and the denoised image obtained after denoising the original image, and obtains the first noise image. Furthermore, by analyzing the number of non-zero pixels and the sum of the pixel values of each pixel point in the first noise, the noise evaluation of the original image can be realized, and the noise score of the original image can be obtained. That is to say, this application uses the denoising result of the original image to reverse-detect the noise of the original image, accurately locates the noise distribution of the original image, while avoiding subjective noise annotation, saving labor costs, and can achieve efficient and accurate noise evaluation.
[0160] Those skilled in the art will readily conceive of other implementations of the embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The embodiments of the present invention are intended to cover any variations, uses, or adaptations of the embodiments of the present invention, which follow the general principles of the embodiments of the present invention and include known common knowledge or conventional technical means in the technical field not disclosed in the embodiments of the present invention. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the embodiments of the present invention are pointed out by the following claims.
[0161] It should be understood that the embodiments of the present invention are not limited to the exact structures already described and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the embodiments of the present invention is only limited by the appended claims.
Claims
1. An image noise evaluation method, characterized in that, Including: Obtain the original image and the denoised image corresponding to the original image, where the denoised image is obtained after denoising the original image; Compare the denoised image with the original image to obtain a first noise image, where the first noise image is used to indicate the noise in the original image; Take the number of non-zero pixels in the first noise image as the first quantity, and take the sum of the pixel values of the pixel points in the first noise image as the second quantity. Based on the ratio of the second quantity to the first quantity, calculate the noise score of the original image, where the noise score is used to evaluate the noise intensity of the original image.
2. The method according to claim 1, wherein The comparing the denoised image with the original image to obtain a first noise image includes: Calculate the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain a second noise image; Extract the texture information of the denoised image to obtain a texture image; In the second noise image, set the pixel values at the corresponding positions of the non-zero pixels in the texture image to 0 to obtain a first noise image.
3. The method according to claim 2, characterized in that The calculating the difference between the pixel values of each corresponding pixel of the denoised image and the original image to obtain a second noise image includes: Calculate the difference between the pixel values of the corresponding same-position pixels in the denoised image and the original image; Determine the absolute value of the difference to obtain a second noise image.
4. The method according to claim 2, wherein The setting the pixel values at the corresponding positions of the non-zero pixels in the texture image to 0 in the second noise image to obtain a first noise image includes: Set the pixels with pixel values less than a first threshold in the texture image to 0 to obtain a reference image, and determine the target positions where the non-zero pixels in the reference image are located; In the second noise image, set the pixel values at the target positions to 0 to obtain a first noise image.
5. The method according to claim 1, wherein The obtaining the original image and the denoised image corresponding to the original image includes: Obtain the original image; Input the original image into a pre-trained denoising model to obtain a denoised image; Wherein, the denoising model is trained by the following steps: Obtain a filtered image, and add noise to the filtered image to obtain an input image, where the filtered image is obtained by filtering a training image; Input the input image into a preset deep learning model for denoising processing to obtain an output image; Calculate the loss value between the output image and the filtered image. When the loss value does not meet the preset threshold, adjust the model parameters of the preset deep learning model until the loss value meets the preset threshold, and take the adjusted preset deep learning model as the denoising model.
6. The method according to claim 1, wherein The determining the noise score of the original image based on the ratio of the second quantity to the first quantity includes: Divide the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter; the preset value is greater than 1; When the value of the candidate parameter is greater than or equal to 1, set the noise score of the original image to 1; When the value of the candidate parameter is less than 1, take the candidate parameter as the noise score of the original image.
7. The method according to claim 6, wherein Before dividing the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter, the method further includes: Determining whether the first quantity is greater than the product of the total number of pixel values in the first noise image and a preset ratio; If it is less than the product, setting the noise score of the original image to 0; If it is not less than the product, performing the step of dividing the ratio of the second quantity to the first quantity by a preset value to obtain a candidate parameter.
8. An image noise evaluation device, characterized in that, The method includes: An acquisition module, configured to acquire an original image and a denoised image corresponding to the original image, where the denoised image is obtained by performing denoising processing on the original image; A comparison module, configured to compare the denoised image with the original image to obtain a first noise image, where the first noise image is used to indicate the noise in the original image; A calculation module, configured to use the number of non-0 pixels in the first noise image as a first quantity, and the sum of the pixel values of the pixel points in the first noise image as a second quantity, and calculate a noise score of the original image based on the ratio of the second quantity to the first quantity, where the noise score is used to evaluate the noise intensity of the original image.
9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, the steps of the image noise evaluation method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the image noise evaluation method according to any one of claims 1 to 7 are implemented.
11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, the image noise evaluation method according to any one of claims 1 to 7 is implemented.