Image quality marking method, device, electronic device and storage medium

Abstract

The present disclosure provides a method, device, electronic device, and storage medium for labeling image quality, and belongs to the field of computer technology. The method includes: obtaining a target image set, the target image set including multiple images; updating the quality parameters of each image in the multiple images, the quality parameters being used to represent the target subject's subjective evaluation of the image quality; in response to the end of the update, for any image, based on the image's quality parameters, adding a quality label to the image, the quality label being used to represent the image quality; wherein, during each update, the multiple images are respectively composed of multiple image pairs based on the real-time quality parameters, the target subject performs subjective evaluations on the multiple image pairs, the difference between the quality parameters of the two images in the image pair is less than a difference threshold, and the images in the multiple image pairs are different from each other. The above scheme can improve the efficiency of adding quality labels to images.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method, device, electronic device, and storage medium for labeling image quality. Background Art

[0002] With the development of computer technology, the scale of images on the Internet is also increasing. During the process of image acquisition, compression, transmission, and display, images may be distorted, resulting in reduced image quality. Therefore, how to accurately annotate image quality is a problem that needs to be solved.

[0003] In related technologies, professionals are usually used as target subjects to rate images, and the image scores are then used to label the image quality. For example, using the Absolute Category Rating (ACR) method, the images to be rated are shown to the target subjects one by one. The target subjects then rate the images on a 5-point scale, and the average score is labeled as the image quality parameter. Among them, 1 point indicates very poor image quality, 2 points indicates poor image quality, 3 points indicates average image quality, 4 points indicates good image quality, and 5 points indicates excellent image quality.

[0004] However, the above method requires the target object to score a large number of images, which has high labor costs and leads to low efficiency in adding quality annotations to images. Summary of the Invention

[0005] The present disclosure provides a method, device, electronic device, and storage medium for labeling image quality, which can improve the efficiency of adding quality labels to images. The technical solutions of the present disclosure are as follows:

[0006] According to one aspect of an embodiment of the present disclosure, a method for labeling image quality is provided, comprising:

[0007] Acquire a target image set, where the target image set includes a plurality of images;

[0008] Updating a quality parameter of each of the plurality of images, wherein the quality parameter is used to represent a target subject's subjective evaluation of the quality of the image;

[0009] In response to the completion of the updating, for any image, adding a quality label to the image based on the quality parameter of the image, wherein the quality label is used to indicate the quality of the image;

[0010] Wherein, each time an update is performed, the multiple images respectively constitute multiple image pairs based on real-time quality parameters, and the target object respectively performs subjective evaluation on the multiple image pairs. The difference between the quality parameters of two images in the image pair is less than the difference threshold, and the images in the multiple image pairs are different from each other.

[0011] In some embodiments, the updating of the quality parameters of each image in the multiple images includes, for the i+1th update, determining multiple image pairs from the target image set based on the quality parameters of each image in the target image set after the i-th update, where i is a positive integer greater than or equal to 1; for any image pair, updating the quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair, the evaluation results being used to indicate the image with higher quality among the two images.

[0012] In some embodiments, for any image pair, based on multiple evaluation results of the target object on the image pair, the quality parameters of the two images in the image pair are updated, including, for any image pair, based on a quality distribution coefficient and the difference between the quality parameters of the two images in the image pair, respectively determining the quality winning probability of the two images, the quality distribution coefficient is used to control the distribution shape of the quality winning probability, and the quality winning probability is used to represent the probability that the quality of one image in the image pair is higher than the quality of the other image; based on the multiple evaluation results and the quality winning probability of the two images, respectively updating the quality parameters of the two images.

[0013] In some embodiments, the quality parameters of the two images are updated respectively based on the multiple evaluation results and the quality winning probabilities of the two images, including, for any image, determining the quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image and the update step size; and updating the quality parameter of the image based on the quality change parameter.

[0014] In some embodiments, for any image, the quality change parameter of the image is determined based on the multiple evaluation results, the quality winning probability of the image and the update step, including determining an average evaluation result based on the evaluation weight of each evaluation result in the multiple evaluation results, the average evaluation result is used to indicate the image with higher quality among the two images, and the evaluation weight is used to represent the confidence level of the evaluation result; for any image, the probability difference between the average evaluation result and the quality winning probability of the image is determined; and the product of the probability difference and the update step is determined as the quality change parameter of the image.

[0015] In some embodiments, the quality change parameter of the image is determined based on the multiple evaluation results, the quality winning probability of the image and the update step size, including, for any evaluation result, obtaining the quality change parameter corresponding to the evaluation result based on the evaluation result and the quality winning probability of the image; and summing the quality change parameters corresponding to the multiple evaluation results to obtain the quality change parameter of the image.

[0016] In some embodiments, for the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, multiple image pairs are determined from the target image set, including for the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, sorting the images to obtain the quality parameter order of the images; forming an image pair from two images whose quality parameter orders are adjacent and whose quality parameter difference is less than the difference threshold, to obtain multiple image pairs, and the images in the multiple image pairs are different from each other.

[0017] In some embodiments, for the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, multiple image pairs are determined from the target image set, including for the (i+1)th update, for any image, based on the quality parameters of each image in the target image set after the (i)th update, determining at least one alternative image whose difference with the quality parameter of the image is less than the difference threshold among multiple images in the target image set that do not constitute an image pair; and forming an image pair with the alternative image having the smallest difference with the quality parameter of the image and the image.

[0018] In some embodiments, in response to the completion of the update, for any image, a quality label is added to the image based on the quality parameters of the image, including in response to the completion of the update, for any image, based on the current quality parameters of the image and the quality parameters of the image after each update, determining the average quality parameter of the image; and adding a quality label to the image based on the average quality parameter.

[0019] In some embodiments, the method further includes initializing the quality parameters of each image in the target image set to initial quality parameters; randomly determining multiple image pairs from the target image set; for any image pair, updating the initial quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair; in response to the completion of the update of the initial quality parameters of each image, determining the quality parameters of each image in the target image set after the first update.

[0020] In some embodiments, the method further includes, for any newly added image, based on the quality parameters of the newly added image after the jth update, determining a first image to which a quality label has been added from the target image set, the first image being used to form a first image pair with the newly added image, the difference between the quality parameters of the first image and the quality parameters of the newly added image being less than the difference threshold, and j being a positive integer greater than or equal to 1; updating the quality parameters of the newly added image based on multiple evaluation results of the target object on the first image pair; and adding a quality label to the newly added image based on the quality parameters of the newly added image in response to the completion of the update.

[0021] In some embodiments, before determining the first image with an annotated quality annotation from the target image set based on the quality parameter of the newly added image after the j-th update, the method also includes initializing the quality parameter of the newly added image to an initial quality parameter; randomly determining a second image from the target image set, the second image being used to form a second image pair with the newly added image; and updating the initial quality parameter of the newly added image based on multiple evaluation results of the target object on the second image pair to obtain the quality parameter of the newly added image after the first update.

[0022] According to another aspect of an embodiment of the present disclosure, a device for labeling image quality is provided, comprising:

[0023] an acquisition unit configured to acquire a target image set, wherein the target image set includes a plurality of images;

[0024] an updating unit configured to update a quality parameter of each of the plurality of images, wherein the quality parameter is used to represent a target subject's subjective evaluation of the quality of the image;

[0025] a labeling unit configured to, in response to completion of the updating, add a quality label to any image based on a quality parameter of the image, wherein the quality label is used to indicate the quality of the image;

[0026] In which, each time an update is performed, the multiple images respectively constitute multiple image pairs based on real-time quality parameters, and the target object respectively performs subjective evaluation on the multiple image pairs. The difference between the quality parameters of two images in the image pair is less than the difference threshold, and the images in the multiple image pairs are different from each other.

[0027] In some embodiments, the updating unit includes:

[0028] a determining subunit configured to, for an i+1th update, determine a plurality of image pairs from the target image set based on quality parameters of each image in the target image set after the i-th update, where i is a positive integer greater than or equal to 1;

[0029] The updating subunit is configured to update, for any image pair, quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair, wherein the evaluation results are used to indicate the image with higher quality among the two images.

[0030] In some embodiments, the update subunit is configured to determine, for any image pair, the quality winning probability of the two images based on the quality distribution coefficient and the difference between the quality parameters of the two images in the image pair, respectively, wherein the quality distribution coefficient is used to control the distribution shape of the quality winning probability, and the quality winning probability is used to represent the probability that the quality of one image in the image pair is higher than the quality of the other image; and based on the multiple evaluation results and the quality winning probability of the two images, respectively update the quality parameters of the two images.

[0031] In some embodiments, the updating subunit includes:

[0032] a determination sub-subunit configured to determine, for any image, a quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image, and the update step size;

[0033] The updating sub-subunit is configured to update the quality parameter of the image based on the quality change parameter.

[0034] In some embodiments, the determination sub-subunit is configured to determine an average evaluation result based on the evaluation weight of each evaluation result in the multiple evaluation results, the average evaluation result is used to indicate the image with higher quality among the two images, and the evaluation weight is used to represent the confidence level of the evaluation result; for any image, determine the probability difference between the average evaluation result and the quality probability of the image; and determine the product of the probability difference and the update step size as the quality change parameter of the image.

[0035] In some embodiments, the determination sub-subunit is configured to obtain, for any evaluation result, a quality change parameter corresponding to the evaluation result based on the evaluation result and the quality winning probability of the image; and sum the quality change parameters corresponding to the multiple evaluation results to obtain the quality change parameter of the image.

[0036] In some embodiments, the determination subunit is configured to, for the i+1th update, sort the images based on the quality parameters of the images in the target image set after the i-th update to obtain the quality parameter order of the images; form an image pair from two images whose quality parameter orders are adjacent and whose quality parameter difference is less than the difference threshold to obtain multiple image pairs, and the images in the multiple image pairs are different from each other.

[0037] In some embodiments, the determination subunit is configured to, for any image, for the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, determine at least one alternative image among multiple images in the target image set that do not constitute an image pair, whose difference in quality parameter with the image is less than the difference threshold; and form an image pair with the alternative image having the smallest difference in quality parameter with the image among the at least one alternative image and the image.

[0038] In some embodiments, the labeling unit is configured to, in response to the completion of the update, determine, for any image, an average quality parameter of the image based on the current quality parameter of the image and the quality parameter of the image after each update; and add a quality label to the image based on the average quality parameter.

[0039] In some embodiments, the updating unit is configured to initialize the quality parameters of each image in the target image set to initial quality parameters; randomly determine multiple image pairs from the target image set; for any image pair, update the initial quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair; in response to the completion of the update of the initial quality parameters of each image, determine the quality parameters of each image in the target image set after the first update.

[0040] In some embodiments, the update unit is configured to determine, for any newly added image, a first image to which a quality label has been added from the target image set based on the quality parameters of the newly added image after the jth update, the first image being used to form a first image pair with the newly added image, the difference between the quality parameters of the first image and the quality parameters of the newly added image being less than the difference threshold, and j being a positive integer greater than or equal to 1; update the quality parameters of the newly added image based on multiple evaluation results of the target object on the first image pair; and add a quality label to the newly added image based on the quality parameters of the newly added image in response to the completion of the update.

[0041] In some embodiments, the updating unit is configured to initialize the quality parameters of the newly added image to the initial quality parameters; randomly determine a second image from the target image set, and the second image is used to form a second image pair with the newly added image; based on multiple evaluation results of the target object on the second image pair, update the initial quality parameters of the newly added image to obtain the quality parameters of the newly added image after the first update.

[0042] According to another aspect of an embodiment of the present disclosure, an electronic device is provided, the electronic device including:

[0043] one or more processors;

[0044] a memory for storing program codes executable by the processor;

[0045] The processor is configured to execute the program code to implement the above-mentioned image quality marking method.

[0046] According to another aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When program code in the computer-readable storage medium is executed by a processor of an electronic device, the electronic device is enabled to perform the above-mentioned image quality labeling method.

[0047] According to another aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program / instruction, which implements the above-mentioned image quality labeling method when executed by a processor.

[0048] The disclosed embodiments provide an image quality annotation scheme. At each update, based on the real-time quality parameters of each image, two images with similar quality parameters are identified from a target image set as an image pair. This generates multiple image pairs, significantly reducing the number of image pairs compared to schemes that pair each image with every other image. This further reduces the workload of the target subject in subjectively evaluating multiple image pairs, lowering labor costs, improving the accuracy of evaluation results, and ultimately increasing the efficiency of adding quality annotations to images.

[0049] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute an improper limitation of the present disclosure.

[0051] Figure 1 The figure is a schematic diagram of an implementation environment according to an exemplary embodiment.

[0052] Figure 2 The figure is a flowchart of a method for marking image quality according to an exemplary embodiment.

[0053] Figure 3 The figure is a flowchart of another method for marking image quality according to an exemplary embodiment.

[0054] Figure 4 The figure is a flowchart of image quality labeling according to an exemplary embodiment.

[0055] Figure 5 is a schematic diagram showing an image pair according to an exemplary embodiment.

[0056] Figure 6 The figure is a block diagram of a device for marking image quality according to an exemplary embodiment.

[0057] Figure 7 The figure is a block diagram of another apparatus for marking image quality according to an exemplary embodiment.

[0058] Figure 8 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0059] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0060] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate 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. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0061] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, storage, display, etc.), and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the target image set and images to be annotated involved in this disclosure were obtained with full authorization.

[0062] Figure 1 FIG1 is a schematic diagram of an implementation environment according to an exemplary embodiment. Figure 1 , the implementation environment specifically includes: a terminal 101 and a server 102.

[0063] The terminal 101 may be at least one of a smartphone, a smartwatch, a desktop computer, a laptop computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), and a laptop computer. In the disclosed embodiment, an application may be installed and run on the terminal 101, and the application is used to display multiple image pairs and obtain the evaluation results of the target object on the image pairs. The target object can log in to the application through the terminal 101 to evaluate the multiple image pairs. The terminal 101 can be connected to the server 102 via a wireless network or a wired network. The server 102 provides background services for the application.

[0064] Terminal 101 may generally refer to one of multiple terminals. This embodiment uses terminal 101 as an example. Those skilled in the art will appreciate that the number of terminals may be greater or lesser. For example, there may be a few terminals, or dozens, hundreds, or even more. This embodiment does not limit the number or device type of terminals.

[0065] The server 102 is at least one of a single server, multiple servers, a cloud computing platform, and a virtualization center. Optionally, the number of the above servers may be more or less, and the embodiments of the present disclosure do not limit this. Of course, the server 102 may also include other functional servers to provide more comprehensive and diversified services. In some embodiments, the server 102 undertakes the main computing work, and the terminal 101 undertakes the secondary computing work; or, the server 102 undertakes the secondary computing work, and the terminal 101 undertakes the main computing work; or, a distributed computing architecture is used between the server 102 and the terminal 101 for collaborative computing. The server 102 can be connected to the terminal 101 and other terminals via a wireless network or a wired network. Optionally, the number of the above servers may be more or less, and the embodiments of the present disclosure do not limit this.

[0066] Figure 2 FIG. 1 is a flow chart showing a method for marking image quality according to an exemplary embodiment. Figure 2 As shown, the method is executed by an electronic device and includes the following steps:

[0067] In step S201 , the electronic device obtains a target image set, where the target image set includes a plurality of images.

[0068] In an embodiment of the present disclosure, the target image set includes multiple images that are images to which quality labels are to be added. The images to which quality labels are to be added may be images with different contents, or may be source images and images obtained by performing image enhancement or image compression processing on the source images. The quality of each image in the target image set is not consistent, and the quality of the image is affected by multiple indicators such as resolution, color depth, and distortion. Therefore, in order to more accurately quantify the quality of the image, the quality of the image to be quality-labeled may be represented by an initial quality parameter. It should be noted that the initial quality parameter is used as a benchmark parameter for reference and cannot accurately represent the actual quality of the image. Subsequently, it is necessary to continuously update the quality parameter of the image based on the initial quality parameter according to the evaluation result of the target object on the image, until a quality parameter that can accurately represent the actual quality of the image is obtained.

[0069] In step S202, the electronic device updates the quality parameters of each image in the multiple images, where the quality parameters are used to represent the target object's subjective evaluation of the image quality. Each time the image is updated, the multiple images constitute multiple image pairs based on the real-time quality parameters, and the target object performs subjective evaluation on the multiple image pairs. If the difference between the quality parameters of two images in the image pair is less than the difference threshold, the images in the multiple image pairs are different from each other.

[0070] In an embodiment of the present disclosure, the electronic device updates the quality parameters of each image by multiple updates. During the multiple updates, the electronic device updates the quality parameters of each image based on the subjective evaluation of each image by the target object. Schematically, if the quality of an image is high and the image gives the target object a good visual experience, then the target object's subjective evaluation of the quality of the image is also high. Accordingly, the electronic device increases the initial quality parameter of the image during the update process. Conversely, if the quality of an image is low and the image gives the target object a poor visual experience, then the target object's subjective evaluation of the quality of the image is also low. Accordingly, the electronic device reduces the initial quality parameter of the image during the update process.

[0071] In some embodiments, at each update, the electronic device selects two images from the target image set whose quality parameter difference is less than a difference threshold based on the real-time quality parameters of multiple images to form an image pair. The difference threshold can be a preset value, such as 30, 60 or 100, which is not limited in the embodiments of the present disclosure. The electronic device displays the above-mentioned multiple image pairs to the target object one by one, and the target object conducts subjective evaluation on the multiple image pairs respectively. When the target object watches any of the displayed image pairs, the target object selects an image with higher quality in the image pair based on his or her subjective feelings. Therefore, it can be considered that the quality of the selected image is higher than the quality of the unselected image. Accordingly, when the electronic device performs this update, it increases the quality parameter of the selected image and reduces the quality parameter of the unselected image.

[0072] It should be noted that the above-mentioned updating process may be an iterative process. Accordingly, one update is one round of iteration. The electronic device updates the image quality parameter once in each round of iteration.

[0073] In step S203, in response to the completion of the update, for any image, the electronic device adds a quality label to the image based on the quality parameter of the image, where the quality label is used to indicate the quality of the image.

[0074] In an embodiment of the present disclosure, when the quality parameters of each image in the target image set are relatively stable, it indicates that the current quality parameters can accurately represent the quality of the image, the update end condition is met, and the update ends. Relative stability can mean that the quality parameters of each image remain unchanged, or that the order of the quality parameters of each image remains unchanged, which is not limited by the embodiment of the present disclosure. In response to the end of the update, for any image, the electronic device performs post-processing operations on the quality parameters of the image to obtain a quality label that can accurately represent the actual quality of the image. The post-processing operations include, but are not limited to, normalizing the quality parameters of the image and performing weighted averaging on the quality parameters of the image.

[0075] The disclosed embodiments provide a method for labeling image quality. At each update, based on the real-time quality parameters of each image, two images with similar quality parameters are identified from a target image set as an image pair. This generates multiple image pairs, significantly reducing the number of image pairs compared to a solution that pairs each image with every other image. This further reduces the workload of the target subject in subjectively evaluating multiple image pairs, lowering labor costs, improving the accuracy of evaluation results, and ultimately increasing the efficiency of adding quality labels to images.

[0076] In some embodiments, the quality parameters of each image in a plurality of images are updated, including for the i+1th update, based on the quality parameters of each image in the target image set after the i-th update, determining a plurality of image pairs from the target image set, where i is a positive integer greater than or equal to 1; for any image pair, based on a plurality of evaluation results of the target object on the image pair, updating the quality parameters of two images in the image pair, the evaluation results being used to indicate the image with higher quality of the two images.

[0077] In the disclosed embodiments, multiple updates are used to create an image pair based on the quality parameters from the last update. The quality parameters of both images in the pair are then updated based on the target object's evaluation of the image pair. Therefore, by updating the image quality parameters through multiple updates, the image quality parameters can be made increasingly accurate over the course of the updates.

[0078] In some embodiments, for any image pair, based on multiple evaluation results of the target object on the image pair, the quality parameters of the two images in the image pair are updated, including for any image pair, based on the quality distribution coefficient and the difference between the quality parameters of the two images in the image pair, respectively determining the quality winning probability of the two images, the quality distribution coefficient is used to control the distribution shape of the quality winning probability, and the quality winning probability is used to represent the probability that the quality of one image in the image pair is higher than the quality of the other image; based on multiple evaluation results and the quality winning probability of the two images, respectively updating the quality parameters of the two images.

[0079] In the disclosed embodiment, by updating the quality parameters of an image using the evaluation results and the quality winning probability, both the subjective evaluation results of the target object and the theoretical winning probability of the image can be taken into account, thereby updating the quality parameters of the image more accurately.

[0080] In some embodiments, based on multiple evaluation results and the quality winning probability of the two images, the quality parameters of the two images are updated respectively, including for any image, based on multiple evaluation results, the quality winning probability of the image and the update step size, determining the quality change parameter of the image; based on the quality change parameter, updating the quality parameter of the image.

[0081] In the embodiment of the present disclosure, by setting the update step size, when updating the quality parameters of the image, the update amplitude can be controlled, thereby improving the accuracy of the quality parameters of the image.

[0082] In some embodiments, for any image, the quality change parameter of the image is determined based on multiple evaluation results, the quality winning probability of the image, and the update step size, including determining an average evaluation result based on the evaluation weight of each evaluation result in the multiple evaluation results, the average evaluation result is used to indicate the image with higher quality of the two images, and the evaluation weight is used to represent the confidence level of the evaluation result; for any image, the probability difference between the average evaluation result and the quality winning probability of the image is determined; the product of the probability difference and the update step size is determined as the quality change parameter of the image.

[0083] In the embodiment of the present disclosure, by setting evaluation weights for the evaluation results, the evaluation results with higher confidence levels can have a greater impact on the average evaluation results than the evaluation results with lower confidence levels, thereby improving the confidence levels of the evaluation results. Through the evaluation results with higher confidence levels, the quality change parameters of the image can be determined more accurately.

[0084] In some embodiments, the quality change parameters of the image are determined based on multiple evaluation results, the quality winning probability of the image, and the update step size, including, for any evaluation result, obtaining the quality change parameters corresponding to the evaluation result based on the evaluation result and the quality winning probability of the image; and summing the quality change parameters corresponding to multiple evaluation results to obtain the quality change parameters of the image.

[0085] In the embodiment of the present disclosure, by determining the quality change parameter corresponding to each evaluation result, the influence of different evaluation results on the change of the quality parameter can be intuitively represented.

[0086] In some embodiments, for the (i+1)th update, multiple image pairs are determined from the target image set based on the quality parameters of each image in the target image set after the (i)th update, including sorting each image based on the quality parameters of each image in the target image set after the (i)th update to obtain a quality parameter order of each image; two images whose quality parameter orders are adjacent and whose quality parameter difference is less than a difference threshold are formed into an image pair to obtain multiple image pairs, and the images in the multiple image pairs are different from each other.

[0087] In the embodiment of the present disclosure, the difference between the quality parameters of two images whose quality parameters are adjacent in order is small. By first determining the images whose quality parameters are adjacent in order, and then from the images whose quality parameters are adjacent, two images whose quality parameter difference is less than a difference threshold are selected to form an image pair. While ensuring that the quality parameters of the two images in the image pair are similar, the efficiency of determining the image pair can also be improved.

[0088] In some embodiments, for the (i+1)th update, multiple image pairs are determined from the target image set based on the quality parameters of each image in the target image set after the i-th update, including for any image, based on the quality parameters of each image in the target image set after the i-th update, determining at least one alternative image whose difference with the quality parameter of the image is less than a difference threshold among multiple images that do not constitute an image pair in the target image set; and forming an image pair with the alternative image and the image whose difference with the quality parameter of the image is the smallest among the at least one alternative image.

[0089] In the embodiment of the present disclosure, for any image, by determining multiple alternative images whose quality parameter differences with the image are less than a difference threshold, and using the alternative image with the smallest difference to form an image pair for the image, it is possible to ensure that the quality parameter differences of the two images in the formed image pair are closest.

[0090] In some embodiments, in response to the completion of the update, for any image, a quality label is added to the image based on the quality parameters of the image, including in response to the completion of the update, for any image, based on the current quality parameters of the image and the quality parameters of the image after each update, determining the average quality parameter of the image; and adding a quality label to the image based on the average quality parameter.

[0091] In the embodiment of the present disclosure, by determining the average quality parameter of the image, the influence of multiple subjective evaluations of the target object during multiple updates can be reduced, thereby reducing random errors and improving the accuracy of image quality annotation.

[0092] In some embodiments, the method further includes initializing the quality parameters of each image in the target image set to initial quality parameters; randomly determining multiple image pairs from the target image set; for any image pair, updating the initial quality parameters of the two images in the image pair based on multiple evaluation results of the target object on the image pair; in response to the completion of the update of the initial quality parameters of each image, determining the quality parameters of each image in the target image set after the first update.

[0093] In the embodiment of the present disclosure, by initializing the quality parameters of each image to an initial quality parameter and using the initial quality parameter as a reference parameter, during the first update process, the quality parameters of each image are updated based on the initial parameter as a reference, and the evaluation results of the target object on the image can be quantified as the update status of the quality parameter, thereby preliminarily obtaining a quality parameter that can represent the actual quality of each image.

[0094] In some embodiments, the method further includes, for any newly added image, based on the quality parameters of the newly added image after the jth update, determining a first image to which a quality label has been added from the target image set, the first image being used to form a first image pair with the newly added image, the difference between the quality parameters of the first image and the quality parameters of the newly added image being less than a difference threshold, and j being a positive integer greater than or equal to 1; updating the quality parameters of the newly added image based on multiple evaluation results of the target object on the first image pair; and adding a quality label to the newly added image based on the quality parameters of the newly added image in response to the completion of the update.

[0095] In the disclosed embodiment, by constructing image pairs with newly added images and images in the target image set that have been quality-annotated, and updating the quality parameters of the newly added images based on the evaluation results of the target objects on the image pairs, new images can be continuously and dynamically quality-annotated, and the theoretical annotation cost is low, which facilitates the construction of a large-scale image dataset with quality annotations.

[0096] In some embodiments, based on the quality parameters of the newly added image after the jth update, before determining the first image to which the quality annotation has been added from the target image set, the method also includes initializing the quality parameters of the newly added image to the initial quality parameters; randomly determining a second image from the target image set, the second image being used to form a second image pair with the newly added image; and updating the initial quality parameters of the newly added image based on multiple evaluation results of the target object on the second image pair to obtain the quality parameters of the newly added image after the first update.

[0097] In the embodiment of the present disclosure, by initializing the quality parameters of each newly added image to an initial quality parameter and using the initial quality parameter as a benchmark parameter, during the first update process, the quality parameters of each newly added image are updated based on the initial parameter as a benchmark, and the evaluation results of the target object on the newly added image can be quantified as the update status of the quality parameter, thereby preliminarily obtaining a quality parameter that can represent the actual quality of each newly added image.

[0098] above Figure 2 The basic process of the present disclosure is shown below. The image quality labeling scheme provided by the present disclosure is further explained. Figure 3 is a flowchart of another image quality marking method according to an exemplary embodiment, the method is executed by an electronic device, see Figure 3 , the method comprising:

[0099] In step S301 , the electronic device obtains a target image set, where the target image set includes a plurality of images.

[0100] In the disclosed embodiments, the target image set is a dataset of images to be quality-labeled, obtained by collecting and organizing image data from real scenes. The multiple images included in the target image set can be images with different content, or they can be source images and images obtained by enhancing or compressing the source images. The electronic device can obtain the target image set from local storage or from other electronic devices, and the disclosed embodiments do not limit this.

[0101] In step S302, the electronic device initializes the quality parameter of each image in the target image set to an initial quality parameter.

[0102] In the disclosed embodiments, the quality of each image in the target image set is not consistent, and image quality is affected by multiple indicators such as resolution, color depth, and distortion. Therefore, to more accurately quantify and label image quality, image quality can be represented by a quality parameter. The higher the image quality, the better the visual experience of the image to the user, and the higher the image quality parameter; the lower the image quality, the worse the visual experience of the image to the user, and the lower the image quality parameter.

[0103] The electronic device initializes the quality parameters of each image in the target image set to initial quality parameters. The initial quality parameters for each image can be the same or different. These initial quality parameters serve as baseline parameters for reference and do not accurately represent the actual quality of the images. Subsequently, based on these initial quality parameters, the image quality parameters need to be continuously updated based on the target subject's evaluation of the image until a quality parameter that accurately represents the actual quality of the image is obtained.

[0104] For example, if the initial quality parameters are the same, the electronic device initializes the initial quality parameters of each image to 1400 points. Subsequently, based on the 1400 points, the quality parameters of each image are increased or decreased according to the evaluation results until a relatively accurate quality parameter for each image is obtained. Alternatively, if the initial quality parameters differ, the electronic device randomly assigns different initial quality parameters to each image in the target image set within a certain parameter range. For example, the electronic device randomly assigns different initial quality parameters to each image within the range of 1200-1600 points.

[0105] It should be noted that after the electronic device initializes the quality parameters of each image to the initial quality parameters, it needs to perform multiple updates. During each update, the image quality parameters are continuously updated based on the target subject's evaluation results of each image until a relatively accurate quality parameter for the image is obtained. In some embodiments, the update process can be an iterative process. Accordingly, one update constitutes one iteration. The electronic device updates the image quality parameters once during each iteration.

[0106] In step S303 , for the first update, the electronic device randomly determines a plurality of image pairs from the target image set.

[0107] In the disclosed embodiments, the example of the same initial quality parameters is used for illustration. During the first update, the electronic device randomly selects two images from the target image set to form an image pair. This random selection method allows for the determination of multiple image pairs, with each image in the pair being distinct, thus preventing the same image from appearing in two pairs.

[0108] For example, the target image set includes N images. When N is an even number, N / 2 image pairs can be determined; when N is an odd number, (N-1) / 2 image pairs can be determined. Unpaired single images will temporarily not participate in this update. During the next update, the image pair will form an image pair with any image in the target image set.

[0109] In step S304, for any image pair, the electronic device updates the initial quality parameters of the two images in the image pair based on multiple evaluation results of the target object on the image pair, and obtains the quality parameters of each image in the target image set after the first update.

[0110] In an embodiment of the present disclosure, an electronic device displays the aforementioned multiple image pairs one by one to a target subject, who then performs a subjective evaluation of each of the multiple image pairs. When viewing any of the displayed image pairs, the target subject selects the image in the pair with higher quality based on their subjective visual perception. Therefore, it can be considered that the quality of the selected image is higher than the quality of the unselected images. Accordingly, during the first update, the electronic device increases the initial quality parameter of the selected image and decreases the initial quality parameter of the unselected image based on the initial quality parameter, thereby obtaining the quality parameters of the two images in the image pair after the first update. By initializing the quality parameter of each image to the initial quality parameter and using this initial quality parameter as a baseline parameter, and updating the quality parameter of each image based on this initial parameter during the first update, the target subject's evaluation of the image can be quantified as the update of the quality parameter, thereby preliminarily obtaining a quality parameter that can represent the actual quality of each image.

[0111] In step S305 , for the (i+1)th update, the electronic device determines a plurality of image pairs from the target image set based on the quality parameters of each image in the target image set after the (i)th update, where (i) is a positive integer greater than or equal to 1.

[0112] In the disclosed embodiment, during the (i+1)th update, the electronic device obtains quality parameters for each image obtained during the (i)th update. Based on the quality parameters for each image, the electronic device determines multiple image pairs. The images in each image pair are distinct, thereby preventing the same image from appearing in both image pairs.

[0113] In some embodiments, the electronic device can form an image pair from two images with similar quality parameters. During the (i+1)th update, the electronic device sorts the images in the target image set based on their quality parameters after the (i)th update, obtaining a quality parameter ranking for each image. The sorting can be from high to low or from low to high. Accordingly, the image with the highest quality parameter can be ranked first or last. The electronic device forms an image pair from two images with adjacent quality parameters and a quality parameter difference less than a difference threshold, obtaining multiple image pairs, each of which contains different images. The difference threshold can be a preset value, such as 30, 60, or 100, and is not limited in the presently disclosed embodiments. The difference between the quality parameters of two images with adjacent quality parameters is relatively small. By first determining images with adjacent quality parameters and then, from among the adjacent quality parameters, selecting two images with a quality parameter difference less than the difference threshold from the adjacent quality parameters to form an image pair, the efficiency of determining image pairs can be improved while ensuring that the quality parameters of the two images in the pair are similar.

[0114] In some embodiments, for any image, the electronic device is capable of determining an image pair consisting of images. During the (i+1)th update, the electronic device determines, based on the quality parameters of each image in the target image set after the i-th update, at least one candidate image whose quality parameter differs from that of the image by less than a difference threshold from the multiple images in the target image set that do not constitute an image pair. The electronic device forms an image pair with the candidate image with the smallest difference in quality parameter from the at least one candidate image. For example, for any image, if the quality parameter of the image after the i-th update is 1480, the electronic device will select three images with quality parameters of 1472, 1464, and 1496 after the i-th update as candidate images. The differences between the quality parameters of these candidate images and the image are 8, 16, and 16, respectively. The electronic device forms an image pair with the candidate image with the smallest difference, i.e., the candidate image with a quality parameter of 1472. For any image, by determining multiple candidate images whose quality parameter differences with the image are less than a difference threshold, and using the candidate image with the smallest difference as the image to form an image pair, it can be ensured that the quality parameter differences of the two images in the formed image pair are closest.

[0115] In step S306, for any image pair, the electronic device updates the quality parameters of the two images in the image pair based on multiple evaluation results of the target object on the image pair, where the evaluation results are used to indicate the image with higher quality of the two images.

[0116] In the disclosed embodiment, during the (i+1)th update, the electronic device determines multiple image pairs from the target image set and then displays each of these image pairs to the target subject, who then performs a subjective evaluation of each of the multiple image pairs. When viewing any of the displayed image pairs, the target subject selects the higher-quality image in the pair based on their subjective visual perception as the target subject's evaluation result. Therefore, the evaluation result can indicate which image the target subject considers to be of higher quality of the two images.

[0117] Accordingly, during the i+1th update, the electronic device increases the quality parameter of the selected image and decreases the quality parameter of the unselected image based on the quality parameters of the two images after the i-th update, thereby obtaining the quality parameters of the two images after the i+1th update. By implementing multiple updates, during any intermediate update, two images with similar quality parameters can be combined into an image pair based on the quality parameters after the previous update. The quality parameters of both images in the image pair can then be updated based on the target subject's evaluation results for any image pair. Therefore, by updating image quality parameters through multiple updates, the image quality parameters can be increasingly accurate over the course of these updates.

[0118] In some embodiments, for any image pair, the electronic device can update the quality parameters of both images based on the evaluation results and the quality winning probabilities of the two images in the image pair. The electronic device determines the quality winning probabilities of the two images based on the quality distribution coefficient and the difference between the quality parameters of the two images in the image pair. The quality distribution coefficient is used to control the distribution shape of the quality winning probability. The larger the quality distribution coefficient, the flatter the distribution curve of the quality winning probability; the smaller the quality distribution coefficient, the steeper the distribution curve of the quality winning probability. The quality winning probability represents the probability that the quality of one image in the image pair is higher than that of the other image. Updating the image quality parameters using the evaluation results and the quality winning probabilities can take into account both the subjective evaluation results of the target subject and the theoretical winning probabilities of the images, thereby more accurately updating the image quality parameters.

[0119] For example, the electronic device can determine the image pair (I A ,I B ) in image I A The quality winning probability P A＞B and image I B The quality winning probability P B＞A .

[0120]

[0121] Wherein, M is the mass distribution coefficient, and the value of M defaults to 400. It can also be set to other values, such as 200, 300 or 500, according to the distribution requirements of the quality winning probability. This embodiment of the present disclosure does not limit this. A and R B They are image I A and image I B The quality parameter P A＞B For image I A The quality winning probability of image I A The quality is higher than that of image I B The probability of the mass. P B＞A For image I B The quality winning probability of image I B The quality is higher than that of image I A It can be seen that if the image I A and image I B The mass is similar, that is, R A ≈R B , then P A＞B =P B＞A =0.5, which means that the probability of the quality of the two images being superior is approximately equal; if the quality of the two images is significantly different, let R A >>R B , then PA＞B ≈1 and P B＞A ≈0, that is, image I A The quality is higher than that of image I B The probability of the mass is infinitely close to 1.

[0122] During the (i+1)th update, after the target subject subjectively selects the image pair and obtains multiple evaluation results, the electronic device updates the quality parameters of the two images based on the multiple evaluation results of the target subject and the quality winning probability of the two images after the i-th update. The multiple evaluation results are multiple evaluation results obtained by multiple target subjects performing subjective evaluations on the same image pair.

[0123] In some embodiments, for any image, the electronic device determines a quality change parameter based on multiple evaluation results, the probability of the image's quality being superior, and an update step size. The update step size controls the magnitude of the quality parameter update during each update. The electronic device updates the image quality parameter based on the quality change parameter. By setting the update step size, the update magnitude can be controlled during image quality parameter updates, improving the accuracy of the image quality parameter.

[0124] The electronic device can determine the image quality change parameter in the following two ways.

[0125] Method 1: The electronic device determines an average evaluation result based on the evaluation weight of each evaluation result from multiple evaluation results. The average evaluation result is used to indicate the higher quality image of the two images. Since different target subjects have different levels of expertise, their evaluation results for the same image pair have different evaluation weights. The evaluation weight is used to indicate the confidence level of the evaluation result. For target subjects with a higher level of expertise, the confidence level of the evaluation result for that target subject is higher, and the evaluation weight is also higher; for target subjects with a lower level of expertise, the confidence level of the evaluation result for that target subject is also lower, and the evaluation weight is also more moderate. The average evaluation result can be represented by 0 or 1: the average evaluation result of the higher quality image of the two images is 1, and conversely, the average evaluation result of the lower quality image is 0. For any image, the electronic device determines the probability difference between the average evaluation result of that image and the probability of the image having better quality. The electronic device multiplies the probability difference by the update step size to determine the quality change parameter of the image. By setting evaluation weights for the evaluation results, the evaluation results with higher confidence levels can have a greater impact on the average evaluation results than the evaluation results with lower confidence levels, thereby improving the confidence levels of the evaluation results. Through the evaluation results with higher confidence levels, the quality change parameters of the image can be determined more accurately.

[0126] For example, the electronic device can determine the image quality change parameter based on the average evaluation result, the image quality winning probability, and the update step size through the update rule shown in the following (2).

[0127]

[0128] Among them, R A and R B They are image I A and image I B The quality parameter after the i-th update. T is the update step size, and T defaults to 16. During the update process, T can also be set to other values, such as 8, 20, or 32, according to the update speed. This embodiment of the present disclosure does not limit this. A and S B They are image I A and I B The average evaluation result, P A＞B and P B＞A They are image I A and image I B The probability of quality winning is: A The higher the quality of S A =1, S B =0, otherwise, if the indicated image I B The higher the quality of S A =0, S B = 1. For any image, the quality change parameter of the image can be determined by multiplying the difference between the average evaluation result of the image and the quality winning probability by the update step size.

[0129] Method 2: For any evaluation result, the electronic device obtains a quality change parameter corresponding to the evaluation result based on the evaluation result and the image's quality probability of success. The evaluation result represents the target subject's subjective evaluation of the image corresponding to the evaluation result. Therefore, the electronic device also needs to sum the quality change parameters corresponding to multiple evaluation results to obtain the image quality change parameter. By determining the quality change parameter corresponding to each evaluation result, it is possible to intuitively represent the impact of different evaluation results on the change in quality parameters.

[0130] For example, for the instruction image I A The first evaluation result with higher quality and the instruction image I B First, based on the first evaluation result, the electronic device determines that the quality change parameters of the two images corresponding to the first evaluation result are T×(S A1 -P A＞B ) and T×(S B1 -P B＞A). Where S A1 =1, S B1 = 0, then, based on the second evaluation result, the electronic device determines that the quality change parameters of the two images corresponding to the second evaluation result are T×(S A2 -P A＞B ) and T×(S B2 -P B＞A ). Among them, S A2 =0, S B2 =1. The electronic device sums the above quality change parameters and obtains the quality change parameters of the two images as T×(S A1 +S A2 -2*P A＞B ) and T×(S B1 +S B2 -2*P B＞A ).

[0131] In step S307 , in response to the update being completed, for any image, the electronic device determines an average quality parameter of the image based on the current quality parameter of the image and the quality parameter of the image after each update.

[0132] In an embodiment of the present disclosure, when the quality parameters of each image in the target image set are relatively stable, it indicates that the current quality parameters can accurately represent the quality of the image, the update end condition is met, and the update ends. Here, relative stability can mean that the quality parameters of each image remain unchanged, or it can mean that the order of the quality parameters of each image remains unchanged, which is not limited by the embodiment of the present disclosure. In response to the end of the update, for any image, the electronic device determines the average quality parameter of the image based on the current quality parameter of the image and the quality parameter of the image after each update. By determining the average quality parameter of the image, the impact of multiple subjective evaluations of the target object during multiple updates can be reduced, thereby reducing random errors.

[0133] For example, the electronic device can determine the average quality parameter of any image by using the following formula (3).

[0134]

[0135] Among them, K is the total number of updates, For image I A The quality parameter after the i-th update is, For image I A The average quality parameter.

[0136] In step S308, the electronic device adds a quality annotation to the image based on the average quality parameter.

[0137] In the disclosed embodiment, for any image, the electronic device performs post-processing on the average quality parameter of the image to obtain a quality label that accurately represents the actual quality of the image. The post-processing operation includes, but is not limited to, normalizing the quality parameter of the image and performing weighted averaging on the quality parameter of the image.

[0138] For example, the electronic device normalizes the average quality parameter of the image using the following formula (4), scales the average quality parameter to a range of 1-5 points, and obtains the quality label of the image.

[0139]

[0140] in, is the minimum value of the quality parameter of each image in the target image set, MOS is the maximum value of the quality parameters of each image. A For image I A Quality annotation, MOS A The value range is [1,5].

[0141] The image quality annotation method provided by the embodiment of the present disclosure can be applied in the scene of subjective quality evaluation of images. In order to more clearly explain the process of subjective quality evaluation of images, the following is combined with Figure 4 The flowchart of image quality annotation shown in FIG. 1 illustrates the process of adding quality annotations to each image in the target image set.

[0142] like Figure 4 As shown, a large number of images of real scenes are first collected to establish a target image dataset to be evaluated. Then, according to the observation distance, observation angle, and number of target objects specified by the ITU-R Rec.BT.500 standard (a subjective quality evaluation standard for images), an offline subjective evaluation environment is set up, and the quality parameters of each image to be evaluated are initialized to the initial quality parameters. Then, images with similar quality parameters are randomly matched to form image pairs, and the image pairs are displayed to the target objects. The target objects perform subjective evaluation of the observed image pairs and select the image with better quality in each image pair. Then, during multiple updates, the electronic device dynamically adjusts the quality parameters of each image based on the evaluation results. When the quality parameters of each image converge, that is, when the quality parameters of each image are relatively stable, the update ends, and the quality parameters of each image are post-processed to obtain the quality label of each image.

[0143] By using images with similar quality parameters as image pairs to be evaluated, there is no need to repeatedly compare them with other images in the target image set. This not only reduces the number of updates and the annotation cost, but also avoids the problem of large differences in the quality parameters of images with different annotation content but similar quality, thereby improving the accuracy of annotation. Figure 5 The image pair shown in the figure has two images with different contents but similar image quality. When the image pair is annotated using the absolute rating scale (ACR), the quality annotations obtained are 3.21 and 4.45, respectively, which is quite different. However, using the method provided by the embodiment of the present disclosure, the quality annotations obtained are 4.75 and 4.87, respectively, which is relatively small. Therefore, the embodiment of the present disclosure provides an image quality annotation method that can, to a certain extent, avoid the influence of environmental factors and subjective bias of the experimenter, and achieve more accurate quality annotation of images with different contents but similar quality.

[0144] It should be noted that after adding quality labels to each image in the target image set, it is also possible to label newly added images based on the target image set, thereby dynamically supplementing the target image set. The process of adding quality labels to newly added images is described below.

[0145] The electronic device initializes the quality parameters of the newly added image to the initial quality parameters. The electronic device randomly determines a second image from the target image set, and the second image is used to form a second image pair with the newly added image. The electronic device updates the initial quality parameters of the newly added image based on the target object's multiple evaluation results of the second image pair, and obtains the quality parameters of the newly added image after the first update. For any newly added image, the electronic device determines a first image with a quality annotation added from the target image set based on the quality parameters of the newly added image after the jth update, and the first image is used to form a first image pair with the newly added image, and the difference between the quality parameters of the first image and the quality parameters of the newly added image is less than a difference threshold, and j is a positive integer greater than or equal to 1. Wherein, j is used to represent the number of updates for adding the quality annotation to the newly added image. The electronic device updates the quality parameters of the newly added image based on the target object's multiple evaluation results of the first image pair. In response to the completion of the update, the electronic device adds a quality annotation to the newly added image based on the quality parameters of the newly added image. By constructing image pairs between the newly added images and the images in the target image set that have been quality-annotated, and updating the quality parameters of the newly added images based on the evaluation results of the target objects on the image pairs, the quality of new images can be continuously and dynamically annotated, and the theoretical annotation cost is low, which facilitates the construction of a large-scale image dataset with quality annotated images.

[0146] Assume that the number of target objects is N, the total number of updates is K, the number of subjective evaluations required for each image during each update is M, and the total number of images in the target image set is B. For any newly added image, the theoretical annotation cost is N×K×M. Using the absolute rating (ACR) method, according to the ITU-R BT.500-11 standard, N>15. Since the target object in ACR directly gives an absolute rating and does not construct an image pair, each image only needs to be subjectively evaluated once during an update, so K=M=1, and the theoretical annotation cost of each newly added image is N. However, using the method provided by the embodiment of the present disclosure, N is usually less than 5, and the number of subjective evaluations required for each image during each update is once, M=1. Therefore, the theoretical annotation cost K of each newly added image is independent of the data size B of the target image set. Among them, the electronic device can determine the maximum total number of updates K by the following formula (5).

[0147]

[0148] in, is the maximum value of the quality parameter of each image, and T is the update step size. That is, the maximum total update number K is the number of times the quality parameter of the newly added image increases from the initial quality parameter of 1400 to the maximum quality parameter of the target image set. K is generally not more than 10.

[0149] The disclosed embodiments provide a method for labeling image quality. At each update, based on the real-time quality parameters of each image, two images with similar quality parameters are identified from a target image set as an image pair. This generates multiple image pairs, significantly reducing the number of image pairs compared to a solution that pairs each image with every other image. This further reduces the workload of the target subject in subjectively evaluating multiple image pairs, lowering labor costs, improving the accuracy of evaluation results, and ultimately increasing the efficiency of adding quality labels to images.

[0150] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present disclosure, and will not be described in detail here.

[0151] Figure 6 FIG. 1 is a block diagram of an image quality labeling device according to an exemplary embodiment. Figure 6 As shown, the device includes: an acquisition unit 601, an update unit 602 and a labeling unit 603.

[0152] An acquiring unit 601 is configured to acquire a target image set, where the target image set includes a plurality of images;

[0153] An updating unit 602 is configured to update a quality parameter of each of the plurality of images, where the quality parameter is used to represent a target subject's subjective evaluation of the quality of the image;

[0154] The labeling unit 603 is configured to, in response to the completion of the updating, add a quality label to any image based on the quality parameter of the image, wherein the quality label is used to indicate the quality of the image;

[0155] Wherein, each time an update is performed, the multiple images respectively constitute multiple image pairs based on real-time quality parameters, and the target object respectively performs subjective evaluation on the multiple image pairs. The difference between the quality parameters of two images in the image pair is less than the difference threshold, and the images in the multiple image pairs are different from each other.

[0156] In some embodiments, Figure 7 is a block diagram of another image quality marking device according to an exemplary embodiment. Figure 7 As shown, the updating unit 602 includes:

[0157] a determining subunit 6021 configured to, for an i+1th update, determine a plurality of image pairs from the target image set based on quality parameters of each image in the target image set after the i-th update, where i is a positive integer greater than or equal to 1;

[0158] The updating subunit 6022 is configured to update, for any image pair, the quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair, wherein the evaluation results are used to indicate the image with higher quality among the two images.

[0159] In some embodiments, the update subunit 6022 is configured to determine, for any image pair, the quality winning probabilities of the two images based on the quality distribution coefficient and the difference between the quality parameters of the two images in the image pair, respectively, wherein the quality distribution coefficient is used to control the distribution shape of the quality winning probability, and the quality winning probability is used to represent the probability that the quality of one image in the image pair is higher than the quality of the other image; and update the quality parameters of the two images based on the multiple evaluation results and the quality winning probabilities of the two images.

[0160] In some embodiments, as Figure 7 As shown, the updating subunit 6022 includes:

[0161] The determination sub-subunit 60221 is configured to determine, for any image, a quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image, and the update step size;

[0162] The updating sub-subunit 60222 is configured to update the quality parameter of the image based on the quality change parameter.

[0163] In some embodiments, the determination sub-subunit 60221 is configured to determine an average evaluation result based on the evaluation weight of each evaluation result in the multiple evaluation results, the average evaluation result is used to indicate the image with higher quality among the two images, and the evaluation weight is used to represent the confidence level of the evaluation result; for any image, determine the probability difference between the average evaluation result and the quality winning probability of the image; and determine the product of the probability difference and the update step size as the quality change parameter of the image.

[0164] In some embodiments, the determination sub-subunit 60221 is configured to obtain, for any evaluation result, a quality change parameter corresponding to the evaluation result based on the evaluation result and the quality winning probability of the image; and sum the quality change parameters corresponding to the multiple evaluation results to obtain the quality change parameter of the image.

[0165] In some embodiments, the determination subunit 6021 is configured to, for the (i+1)th update, sort the images based on the quality parameters of the images in the target image set after the (i)th update to obtain an order of quality parameters of the images; form an image pair from two images whose quality parameter orders are adjacent and whose quality parameter difference is less than the difference threshold to obtain multiple image pairs, and the images in the multiple image pairs are different from each other.

[0166] In some embodiments, the determination subunit 6021 is configured to, for any image, for the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, determine at least one alternative image whose difference with the quality parameter of the image is less than the difference threshold among multiple images in the target image set that do not constitute an image pair; and form an image pair with the alternative image having the smallest difference with the quality parameter of the image among the at least one alternative image and the image.

[0167] In some embodiments, the labeling unit 603 is configured to, in response to the completion of the update, determine, for any image, an average quality parameter of the image based on the current quality parameter of the image and the quality parameter of the image after each update; and add a quality label to the image based on the average quality parameter.

[0168] In some embodiments, the updating unit 602 is configured to initialize the quality parameters of each image in the target image set to initial quality parameters; randomly determine multiple image pairs from the target image set; for any image pair, based on multiple evaluation results of the target object on the image pair, update the initial quality parameters of two images in the image pair; in response to the completion of the update of the initial quality parameters of each image, determine the quality parameters of each image in the target image set after the first update.

[0169] In some embodiments, the updating unit 602 is configured to determine, for any newly added image, a first image to which a quality annotation has been added from the target image set based on the quality parameters of the newly added image after the jth update, the first image being used to form a first image pair with the newly added image, the difference between the quality parameters of the first image and the quality parameters of the newly added image being less than the difference threshold, and j being a positive integer greater than or equal to 1; the quality parameters of the newly added image are updated based on multiple evaluation results of the target object on the first image pair; the annotation unit 603 is configured to add a quality annotation to the newly added image based on the quality parameters of the newly added image in response to the end of the update.

[0170] In some embodiments, the updating unit 602 is configured to initialize the quality parameters of the newly added image to the initial quality parameters; randomly determine a second image from the target image set, and the second image is used to form a second image pair with the newly added image; based on multiple evaluation results of the target object on the second image pair, update the initial quality parameters of the newly added image to obtain the quality parameters of the newly added image after the first update.

[0171] The disclosed embodiments provide an image quality annotation device. During each update, based on the real-time quality parameters of each image, two images with similar quality parameters are identified from a target image set as an image pair. This generates multiple image pairs, significantly reducing the number of image pairs compared to a solution that pairs each image with every other image. This further reduces the workload of the target subject in subjectively evaluating multiple image pairs, lowering labor costs, improving the accuracy of evaluation results, and ultimately increasing the efficiency of adding quality annotations to images.

[0172] It should be noted that the image quality annotation device provided in the above embodiment is merely an example of the division of the above functional units. In actual applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the electronic device can be divided into different functional units to complete all or part of the functions described above. In addition, the image quality annotation device provided in the above embodiment and the image quality annotation method embodiment are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0173] Regarding the image quality labeling device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0174] Figure 8 8 is a block diagram of an electronic device according to an exemplary embodiment. Generally, the electronic device 800 includes a processor 801 and a memory 802 .

[0175] The processor 801 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 801 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 801 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 801 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 801 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0176] The memory 802 may include one or more computer-readable storage media, which may be non-transitory. The memory 802 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 802 is used to store at least one program code, which is executed by the processor 801 to implement the image quality labeling method provided in the method embodiment of the present disclosure.

[0177] In some embodiments, electronic device 800 may optionally include a peripheral device interface 803 and at least one peripheral device. The processor 801, memory 802, and peripheral device interface 803 may be connected via a bus or signal lines. Each peripheral device may be connected to peripheral device interface 803 via a bus, signal lines, or circuit boards. Specifically, the peripheral device may include at least one of a radio frequency circuit 804, a display screen 805, a camera assembly 806, an audio circuit 807, and a power supply 808.

[0178] The peripheral device interface 803 can be used to connect at least one I / O (Input / Output)-related peripheral device to the processor 801 and the memory 802. In some embodiments, the processor 801, the memory 802, and the peripheral device interface 803 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 801, the memory 802, and the peripheral device interface 803 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0179] The RF circuit 804 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 804 communicates with communication networks and other communication devices via electromagnetic signals. The RF circuit 804 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the RF circuit 804 includes an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, and the like. The RF circuit 804 can communicate with other electronic devices via at least one wireless communication protocol. Such wireless communication protocols include, but are not limited to, metropolitan area networks (MANs), various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks (WLANs), and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 804 may also include circuits related to NFC (Near Field Communication), which is not limited in this disclosure.

[0180] Display screen 805 is used to display a user interface (UI). This UI can include graphics, text, icons, videos, or any combination thereof. When display screen 805 is a touch screen display, it can also capture touch signals on or above the surface of display screen 805. These touch signals can be input as control signals to processor 801 for processing. Display screen 805 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there can be a single display screen 805, located on the front panel of electronic device 800. In other embodiments, there can be at least two display screens 805, located on different surfaces of electronic device 800 or in a foldable design. In still other embodiments, display screen 805 can be a flexible display, located on a curved or foldable surface of electronic device 800. Display screen 805 can also be configured as a non-rectangular, irregular shape, also known as a special-shaped screen. Display screen 805 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0181] The camera assembly 806 is used to capture images or videos. Optionally, the camera assembly 806 includes a front camera and a rear camera. Typically, the front camera is arranged on the front panel of the electronic device, and the rear camera is arranged on the back of the electronic device. In some embodiments, there are at least two rear cameras, which are any one of a main camera, a depth of field camera, a wide-angle camera, and a telephoto camera, so as to realize the fusion of the main camera and the depth of field camera to realize the background blur function, the fusion of the main camera and the wide-angle camera to realize panoramic shooting and VR (Virtual Reality) shooting function or other fusion shooting functions. In some embodiments, the camera assembly 806 may also include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cold light flash, which can be used for light compensation at different color temperatures.

[0182] The audio circuit 807 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals to be input into the processor 801 for processing, or input into the radio frequency circuit 804 to achieve voice communication. For the purpose of stereo sound collection or noise reduction, there can be multiple microphones, which are respectively arranged in different parts of the electronic device 800. The microphone can also be an array microphone or an omnidirectional collection microphone. The speaker is used to convert the electrical signals from the processor 801 or the radio frequency circuit 804 into sound waves. The speaker can be a traditional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signals into sound waves audible to humans, but also convert the electrical signals into sound waves inaudible to humans for purposes such as ranging. In some embodiments, the audio circuit 807 may also include a headphone jack.

[0183] Power supply 808 is used to power the various components of electronic device 800. Power supply 808 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 808 includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0184] Those skilled in the art will understand that Figure 8 The structure shown in the figure does not constitute a limitation on the electronic device 800, and the electronic device 800 may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0185] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 802 including instructions. The instructions can be executed by the processor 801 of the terminal 800 to implement the above-described image quality labeling method. Alternatively, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0186] A computer program product includes a computer program, which implements the above-mentioned image quality marking method when executed by a processor.

[0187] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0188] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for marking image quality, characterized in that: The method comprises: Acquire a target image set, where the target image set includes a plurality of images; Updating a quality parameter of each of the plurality of images, wherein the quality parameter is used to represent a target subject's subjective evaluation of the quality of the image; In response to the completion of the updating, for any image, adding a quality label to the image based on the quality parameter of the image, wherein the quality label is used to indicate the quality of the image; Wherein, each time an update is performed, the multiple images respectively constitute multiple image pairs based on real-time quality parameters, and the target object respectively performs subjective evaluation on the multiple image pairs. The difference between the quality parameters of two images in the image pair is less than the difference threshold, and the images in the multiple image pairs are different from each other.

2. The image quality labeling method according to claim 1, characterized in that: Updating the quality parameter of each image in the plurality of images includes: For the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, determining a plurality of image pairs from the target image set, where (i) is a positive integer greater than or equal to 1; For any image pair, quality parameters of two images in the image pair are updated based on multiple evaluation results of the target object on the image pair, where the evaluation results are used to indicate the image with higher quality among the two images.

3. The image quality labeling method according to claim 2, characterized in that: For any image pair, updating the quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair includes: For any image pair, determining respectively the quality winning probabilities of the two images based on the quality distribution coefficient and the difference between the quality parameters of the two images in the image pair, wherein the quality distribution coefficient is used to control the distribution shape of the quality winning probabilities, and the quality winning probabilities are used to represent the probability that the quality of one image in the image pair is higher than the quality of the other image; Based on the multiple evaluation results and the quality winning probabilities of the two images, the quality parameters of the two images are updated respectively.

4. The image quality labeling method according to claim 3, characterized in that: The updating of the quality parameters of the two images based on the multiple evaluation results and the quality winning probabilities of the two images respectively includes: For any image, determining a quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image, and the update step size; Based on the quality change parameter, a quality parameter of the image is updated.

5. The image quality labeling method according to claim 4, characterized in that: The step of determining, for any image, a quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image, and the update step size includes: determining an average evaluation result based on an evaluation weight of each evaluation result in the plurality of evaluation results, wherein the average evaluation result is used to indicate the image with higher quality among the two images, and the evaluation weight is used to indicate a confidence level of the evaluation result; For any image, determining a probability difference between the average evaluation result and the quality winning probability of the image; The product of the probability difference and the update step size is determined as the quality change parameter of the image.

6. The image quality labeling method according to claim 4, characterized in that: The determining of the quality change parameter of the image based on the multiple evaluation results, the quality winning probability of the image, and the update step size includes: For any evaluation result, based on the evaluation result and the quality winning probability of the image, a quality change parameter corresponding to the evaluation result is obtained; The quality change parameters corresponding to the multiple evaluation results are summed to obtain the quality change parameter of the image.

7. The image quality labeling method according to claim 2, characterized in that: For the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, determining a plurality of image pairs from the target image set comprises: For the (i+1)th update, sorting the images based on the quality parameters of the images in the target image set after the (i)th update to obtain a quality parameter order of the images; Two images whose quality parameters are adjacent in order and whose quality parameter difference is less than the difference threshold form an image pair, thereby obtaining a plurality of image pairs, wherein the images in the plurality of image pairs are different from each other.

8. The image quality labeling method according to claim 2, characterized in that: For the (i+1)th update, based on the quality parameters of each image in the target image set after the (i)th update, determining a plurality of image pairs from the target image set comprises: For the (i+1)th update, for any image, based on the quality parameters of each image in the target image set after the (i)th update, determine at least one candidate image from a plurality of images in the target image set that do not constitute an image pair, the difference between the quality parameter of the image and that of the image being described is less than the difference threshold. An image pair is formed by combining the candidate image having the smallest difference between the quality parameter of the at least one candidate image and the image.

9. The image quality labeling method according to claim 1, characterized in that: In response to the completion of the updating, for any image, adding a quality annotation to the image based on the quality parameter of the image includes: In response to the update being completed, for any image, determining an average quality parameter of the image based on the current quality parameter of the image and the quality parameter of the image after each update; A quality annotation is added to the image based on the average quality parameter.

10. The image quality labeling method according to claim 1, characterized in that: The method further comprises: Initializing the quality parameter of each image in the target image set to an initial quality parameter; randomly determining a plurality of image pairs from the target image set; For any image pair, updating initial quality parameters of two images in the image pair based on multiple evaluation results of the target object on the image pair; In response to the completion of the updating of the initial quality parameters of the respective images, the quality parameters of the respective images in the target image set after the first update are determined.

11. The image quality labeling method according to claim 1, characterized in that: The method further comprises: For any newly added image, based on the quality parameter of the newly added image after the j-th update, determine a first image to which a quality annotation has been added from the target image set, the first image being used to form a first image pair with the newly added image, the difference between the quality parameter of the first image and the quality parameter of the newly added image being less than the difference threshold, where j is a positive integer greater than or equal to 1; updating the quality parameter of the newly added image based on a plurality of evaluation results of the target object on the first image pair; In response to the completion of the update, a quality label is added to the newly added image based on the quality parameter of the newly added image.

12. The image quality labeling method according to claim 11, characterized in that: Before determining the first image to which the quality annotation has been added from the target image set based on the quality parameter of the newly added image after the j-th update, the method further includes: Initializing the quality parameter of the newly added image to the initial quality parameter; randomly determining a second image from the target image set, where the second image is used to form a second image pair with the newly added image; Based on multiple evaluation results of the target object on the second image pair, the initial quality parameter of the newly added image is updated to obtain the quality parameter of the newly added image after the first update.

13. An image quality labeling device, characterized in that: The device comprises: an acquisition unit configured to acquire a target image set, wherein the target image set includes a plurality of images; an updating unit configured to update a quality parameter of each of the plurality of images, wherein the quality parameter is used to represent a target subject's subjective evaluation of the quality of the image; a labeling unit configured to, in response to completion of the updating, add a quality label to any image based on a quality parameter of the image, wherein the quality label is used to indicate the quality of the image; Wherein, each time an update is performed, the multiple images respectively constitute multiple image pairs based on real-time quality parameters, and the target object respectively performs subjective evaluation on the multiple image pairs. The difference between the quality parameters of two images in the image pair is less than the difference threshold, and the images in the multiple image pairs are different from each other.

14. An electronic device, characterized in that: The electronic device comprises: one or more processors; a memory for storing program code executable by the processor; The processor is configured to execute the program code to implement the image quality marking method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the image quality labeling method according to any one of claims 1 to 12.

Images