Methods, devices, equipment, media, and products for selecting image quality levels.

Abstract

This paper relates to a method, apparatus, device, medium, and product for selecting image quality levels. The method includes: obtaining a first request received by a client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; obtaining at least one of network status information and device performance information of the electronic device where the client is located, and obtaining an upper limit for loading time and a lower limit for image quality related to the client; selecting a first quality level suitable for the client from the multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit for loading time, and the lower limit for image quality; and displaying the first image on the client based on the selected first quality level. This paper effectively improves the accuracy of image quality selection results and significantly enhances the user experience.

Description

Technical Field

[0001] This article relates to the field of computer technology, and in particular to a method, apparatus, device, medium and product for selecting image quality levels. Background Technology

[0002] With the continuous development of smart terminals and network technology, images have become an important carrier of media information and one of the main ways for people to obtain information through clients in their daily lives and work. The presentation effect of images on the client largely determines the quality of the user experience. Some related technologies set multiple quality levels for the images to be displayed (e.g., different quality levels have different resolutions) and select the quality level that is suitable for the client so that the image corresponding to that quality level can be displayed to the user on the client. However, the inventors found through research that the quality level selection method used by the related technologies is not good. The quality level selection result is usually not the quality level that is truly suitable for the client, resulting in poor accuracy and a poor user experience. Therefore, there is an urgent need for a new image quality level selection technology. Summary of the Invention

[0003] To address, or at least partially address, the aforementioned technical problems, this paper provides a method, apparatus, device, medium, and product for selecting image quality levels.

[0004] In a first aspect, this paper provides a method for selecting an image quality level. The method includes: obtaining a first request received by a client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; obtaining at least one of network status information and device performance information of the electronic device where the client is located, and obtaining an upper limit for loading time and a lower limit for image quality related to the client; selecting a first quality level adapted to the client from the multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit for loading time, and the lower limit for image quality; and displaying the first image on the client based on the selected first quality level.

[0005] Secondly, this paper also provides an image quality level selection device, comprising: a request receiving module, configured to acquire a first request received by a client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; an information acquisition module, configured to acquire at least one of network status information and device performance information of the electronic device where the client is located, and acquire an upper limit of loading time and a lower limit of image quality related to the client; a quality level selection module, configured to select a first quality level adapted to the client from multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit of loading time, and the lower limit of image quality; and an image display module, configured to display the first image on the client according to the selected first quality level.

[0006] Thirdly, this document also provides an electronic device comprising: a storage device storing a computer program thereon; and a processing device for executing the computer program in the storage device to implement the image quality level selection method provided herein.

[0007] Fourthly, this document also provides a computer-readable storage medium storing a computer program for performing the image quality level selection method provided herein.

[0008] Fifthly, this document also provides a computer program product stored in a computer storage medium and including computer-executable instructions that, when executed by a device, cause the device to perform the image quality level selection method provided herein.

[0009] The technical solution provided in this paper can, upon receiving a first request (request to display a first image), obtain at least one of the network status information and device performance information of the client's electronic device, as well as the upper limit of loading time and the lower limit of image quality related to the client. Then, based on the obtained information, it selects a first quality level suitable for the client from multiple quality levels corresponding to the first image, and displays the first image on the client according to the selected first quality level. This method fully considers that the accuracy of simply selecting a quality level based on a single factor in related technologies is poor, and that the image loading time and image quality corresponding to different quality levels are the main factors affecting user experience. Therefore, it selects a quality level suitable for the client based on multiple reference factors such as network status information, at least one of the device performance information, the upper limit of loading time related to the client, and the lower limit of image quality. This effectively improves the accuracy of the image selection result and helps ensure that the first image displayed on the client can better meet the image quality and loading time requirements, thereby improving the user experience.

[0010] It should be understood that the descriptions in this section are not intended to identify key or essential features of the examples in this document, nor are they intended to limit the scope of this document. Other features of this document will become readily apparent from the following description. Attached Figure Description

[0011] The accompanying drawings, which are incorporated in and form part of this specification, illustrate examples consistent with this document and, together with the specification, serve to explain the principles herein.

[0012] To more clearly illustrate the technical solutions in this document or the prior art, the accompanying drawings used in the examples or prior art descriptions will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0013] Figure 1 This is a schematic diagram illustrating one application scenario provided in this article; Figure 2 A flowchart illustrating a method for selecting image quality levels provided in this paper; Figure 3 This article provides a schematic diagram of the process for selecting image quality levels. Figure 4 This article provides a schematic diagram of the process for selecting image quality levels. Figure 5 This article provides a schematic diagram of the process for selecting image quality levels. Figure 6 A schematic diagram of an image quality level selection device provided in this paper; Figure 7 This is a schematic diagram of the structure of an electronic device provided in this article. Detailed Implementation

[0014] To better understand the objectives, features, and advantages outlined in this paper, the proposed solution will be further described below. It should be noted that, unless otherwise specified, the examples and features described in this paper can be combined with each other.

[0015] It is understandable that before using the technical solutions disclosed in the various embodiments herein, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this document in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0016] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as electronic devices, applications, servers, or storage media, that perform the operations described herein, based on the prompt message.

[0017] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0018] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation method described in this article. Other methods that comply with relevant laws and regulations may also be applied to the implementation method described in this article.

[0019] The following description sets forth many specific details to provide a full understanding of this document, but it may also be implemented in other ways than those described herein; obviously, the examples in the specification are only a part of the examples in this document, and not all of them.

[0020] Most related technologies select image quality levels based on a single factor, simply establishing a fixed mapping relationship between a single factor and quality levels. While simple to implement, the selected quality level is often not the one truly suited to the client. The inventors discovered that the main reasons are: 1) The mobile network environment is complex. Specifically, existing mobile network environments generally exhibit high dynamism and complexity. For example, user devices may frequently switch between different types of cellular networks, Wi-Fi, and other network access methods, with network bandwidth fluctuations reaching several orders of magnitude. This highly dynamic network environment poses a severe challenge to traditional image quality selection technologies. 2) User device performance is significantly stratified. Specifically, there are currently a wide variety of user devices with significant differences in hardware configuration. Differences in CPU (Central Processing Unit) processing power, memory capacity, GPU (Graphics Processing Unit) rendering performance, and display resolution among different devices make it difficult for single-factor image quality selection strategies to achieve optimal performance on all devices. 3) User needs vary. Specifically, different users have significant individual differences in their preferences for image loading speed and display quality. Some users prioritize loading speed, while others value display quality more. A single-factor image selection strategy cannot adequately address these personalized needs. In summary, related technologies make selection decisions based on only a single factor, lacking a comprehensive understanding of multi-dimensional information, resulting in insufficient decision accuracy. Furthermore, these technologies typically focus on a single optimization objective, such as making selection decisions solely based on image loading time or image display quality, failing to establish a multi-objective collaborative optimization approach and unable to achieve a balance between conflicting objectives such as loading time and display quality. Additionally, these technologies employ static decision rules, neglecting users' personalized needs, all of which contribute to the low accuracy of the final image selection results.

[0021] The deficiencies in the related technologies are the result of the inventor's practical experience and careful research. Therefore, the discovery process of the aforementioned deficiencies and the solutions proposed below should be considered as the inventor's contributions to this paper. To improve at least one of the aforementioned problems in the related technologies, this paper provides a method for selecting image levels, which is described in detail below.

[0022] First, to facilitate understanding of the application scenario in this article, for example, refer to the following: Figure 1The diagram illustrates an application scenario, showing the potential client and server components. In some scenarios, users can operate through the client, such as triggering a client page to display media resources like images, viewing the images, and further interacting with them by clicking or saving them. In some scenarios, the image quality level selection method provided in this paper can be executed by the client; in others, it can be executed through interaction between the client and server, or solely by the server, allowing for flexible configuration. This paper does not restrict the client's implementation method; it can be implemented using software and / or hardware, specifically through electronic devices. These electronic devices can be any type of mobile terminal, fixed terminal, or portable terminal, and in some examples, they can support any type of interface. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, security services, content delivery networks, or big data and artificial intelligence platforms. The server can include, for example, computing systems / servers, such as mainframes, edge computing nodes, and computing devices in cloud environments. In some cases, the server can provide background services to the client. A communication connection can be established between the server and the client, which can be established via wired or wireless means. The communication connection can include, but is not limited to, Bluetooth connections, mobile network connections, Universal Serial Bus connections, and Wi-Fi connections, etc., and this document is not limited in this respect.

[0023] Figure 2 This document presents a flowchart illustrating a method for selecting image quality levels. This method can be executed by an image quality level selection device, which can be implemented using software and / or hardware, and is typically integrated into an electronic device. Figure 1 As shown, the method mainly includes the following steps 202 to 208: Step 202: Obtain the first request received by the client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels.

[0024] This document does not restrict the sender of the first request. In some cases, the first request is sent by the client, such as when a user requests the display of the client's first page, which is the page used to display the first image; or when a user interacts with the client's first component, which is the component used to display the first image. In other cases, the first request is sent by the server, such as when the server sends the first request to the client when the timing for displaying the first image is reached, so that the client can display the first image. The timing for displaying the first image can be determined according to a preset strategy or based on the interaction between the user and the client.

[0025] The first image mentioned above is the image to be displayed. This document does not limit the content or number of the first images; there may be one or more first images. The first image corresponds to multiple quality levels, each with different quality level parameters, including but not limited to quality parameters such as resolution. Images obtained by transcoding the same source image at different quality levels will have different qualities. In some cases, quality can be characterized by image quality. In some examples, the first request mentioned above is a request sent by the server. This first request also carries image description information corresponding to the first image at multiple quality levels. The image description information includes, but is not limited to, image resolution and image data size. Furthermore, the image description information may also include information such as the image address. Optionally, the image description information may also include the image quality obtained based on a quality assessment algorithm. Step 204: Obtain at least one of the network status information and device performance information of the electronic device where the client is located, as well as the upper limit of loading time and the lower limit of image quality related to the client.

[0026] This article does not restrict the methods for obtaining network status information and device performance information. For example, the above information can be obtained entirely by the client, obtained by the server, or obtained by both in combination.

[0027] In some cases, the upper limit of loading time and the lower limit of image quality related to the client can be preset values. In other cases, these limits can be determined based on relevant client information. For example, they can be determined based on client interaction information, which is information related to the first interaction record of multiple historically displayed images for the corresponding client. The upper limit of loading time and the lower limit of image quality may differ for different clients, which helps to make more reasonable image selection based on the client's own situation.

[0028] Step 206: Select the first quality level that is compatible with the client from multiple quality levels corresponding to the first image, based on at least one of the network status information and device performance information, the upper limit of loading time, and the lower limit of image quality.

[0029] In some examples, image prediction information for multiple quality levels corresponding to the first image can be further obtained. This image prediction information includes image loading time and image quality. For example, the image prediction information can be determined based on at least one of the aforementioned image description information, network status information, and device performance information. This image prediction information can be used to compare with the upper limit of loading time and the lower limit of image quality in subsequent comparisons, so that the comparison results can be used as a reference for quality level selection. In this case, a first quality level suitable for the client can be selected from the multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the image prediction information for multiple quality levels corresponding to the first image, the upper limit of loading time, and the lower limit of image quality.

[0030] Step 208: Display the first image on the client according to the selected first quality level.

[0031] In some cases, the first image can be displayed directly on the client's interface based on the resolution and other parameters of the selected first quality level. In other cases, the client can request the first image from the server based on the first quality level, and display the first image on the interface upon receiving it.

[0032] The above method fully considers that the accuracy of selecting quality levels based solely on a single factor in related technologies is poor. Since the image loading time and image quality corresponding to different quality levels are the main factors affecting user experience, the method selects a quality level that is compatible with the client based on multiple reference factors such as network status information, device performance information, upper limit of loading time related to the client, and lower limit of image quality. This can effectively improve the accuracy of image selection results and help ensure that the first image displayed by the client can better meet the image quality and loading time requirements, thereby improving the user experience.

[0033] To facilitate understanding, this article provides examples of how to obtain reference information such as network status information, device performance information, and interaction information that may be required in the image selection process. The following sections explain the methods for obtaining network status information, device performance information, and interaction information: (a) Methods for obtaining network status information In some cases, network status information includes network quality assessment results and / or bandwidth prediction results, which will be explained below.

[0034] Optionally, network quality assessment results can be characterized by network type, network quality level, or network quality value. For example, the quality of the network near the client on a single network request link corresponding to the client can be assessed. For instance, if there are multiple network segments between the client and the server, the first segment corresponds to the network link between the client and the public network exit, the second segment corresponds to the network link between the public network exit and the edge node, and the third segment corresponds to the network link between the edge node and the server, then the quality of the first segment can be mainly assessed to more accurately measure the quality of the network status on the client side. In addition, the overall network quality level of the user can be assessed based on the link quality monitoring results, such as the percentage of abnormal links. The above are just examples and should not be regarded as limitations. In practical applications, network quality assessment techniques can be flexibly selected. In some cases, network quality assessment results can be based on network type characterization. Optional network types include no network, fake network, forced portal, weak network, and strong network. Each network type can correspond to a network quality level or network quality value. The no-network type refers to a network environment with no signal, such as Wi-Fi or cellular networks, meaning there is absolutely no network connection. The fake network type refers to a situation where, despite having a network connection, the client does not receive any network response data after sending a request. The forced portal type refers to a situation where, despite having a network connection, the client can receive data packets returned by the server, but because the client has not been authenticated by the gateway, the data is intercepted by the gateway, and the gateway will return a message informing the client that login authentication is required. The weak and strong network types refer to a comprehensive network quality rating based on indicators such as connection establishment time, packet loss rate, retransmission rate, and data transmission speed. This reflects the relative ranking of link latency, packet loss, and bandwidth relative to the overall user experience of the application. If the weighted score of each indicator is below a certain percentile, it is considered a weak network; otherwise, it is considered a strong network. Evaluating network quality level or network quality value based on the above network types can effectively determine the user's current network status. For example, it can promptly change the parameters related to the selection strategy after the user switches from a strong network to a weak network, which helps to select a selection level that is more suitable for the current weak network status and ensures that more images can be loaded and displayed as quickly as possible even in a weak network.

[0035] Optionally, bandwidth prediction results can be characterized by bandwidth prediction values. In modern mobile application scenarios, the network environment is highly dynamic and complex. Mobile applications typically acquire resources through multiple concurrent requests, resulting in network bandwidth being shared by multiple requests. This concurrency characteristic makes it impossible for a single request to exclusively occupy network resources, increasing the complexity of bandwidth measurement. Simultaneously, the wireless network environment is affected by various factors, including signal strength, interference levels, and network congestion, leading to significant temporal fluctuations in bandwidth. Furthermore, images to be displayed on client pages, such as cover images, are typically small in size and require extremely short theoretical transmission times, such as tens of milliseconds. For these small file transfers, network connection establishment time accounts for a significant proportion of the total time, causing a large deviation in the bandwidth value calculated directly using the total time. To improve these issues, this paper provides a bandwidth prediction method that can predict bandwidth based on the download information of historical displayed images on the client. Optionally, it can combine a dynamic weighted average algorithm and a window smoothing method to generate bandwidth prediction values ​​on the client side. For example, assuming the sliding window length is M, and the window contains the most recent N image download sampling points, where N ≤ M, the file size, download request start timestamp, total download request time, and weighted weight corresponding to each image download sampling point can be obtained. Specifically, the network connection time includes the time spent on DNS (Domain Name System) resolution and TCP (Transmission Control Protocol) connection establishment. The weighted weight of each image download sampling point is equal to the difference between 1 and a target ratio. The target ratio is equal to the ratio between the network connection time corresponding to the image download sampling point and the total download request time. Optionally, the current smoothed bandwidth prediction value is equal to the sum of the first product result and the second product result. The first product result is equal to the product of the first coefficient and the weighted average of the bandwidth values ​​of the N image download sampling points. The second product result is equal to the product of the second coefficient and the historical smoothed bandwidth value. The first coefficient can also be called a smoothing factor, and the second coefficient is equal to the difference between 1 and the first coefficient. The bandwidth value of each image download sampling point is obtained based on the ratio between the file size and the total request time. The above method can be used to objectively predict the current bandwidth value by combining historical bandwidth values, and the bandwidth prediction result calculated based on the window smoothing method is smoother and more reliable.

[0036] (ii) Methods for obtaining equipment performance information In some cases, device performance information can be determined based on one or more of the following: the hardware and software configuration information of the client's electronic device; at least one offline test performance score; a performance score collected periodically by the client; and a performance score calculated in real time during client use. The periodically collected performance score includes, but is not limited to, scores obtained based on one or more of the following metrics: CPU performance, memory performance, network performance, throughput-related performance, client startup time, and smoothness-related metrics. The performance score calculated in real time during client use is used to present performance information such as the user's device load and hardware / software decoding performance within a recent time period. Optionally, the performance score can be normalized to the range [0, 100]. A higher performance score indicates better device performance, enabling it to play higher bitrate media content and handle larger-scale multi-threaded tasks. Optionally, at least one offline performance test score corresponding to the client can be weighted to obtain a comprehensive offline performance score, and at least one online performance test score corresponding to the client can be weighted to obtain a comprehensive online performance score. The client's performance score is obtained by weighting the comprehensive offline performance score and the online performance test score. Optionally, the above-mentioned online performance test score can be implemented with reference to the performance score collected by the client periodically and the performance score calculated in real time during the client's use. This paper does not restrict the offline performance test indicators and the online performance test indicators, and they can be flexibly set. Through the above method, the objectivity and reliability of the final client performance score can be guaranteed.

[0037] (III) Methods for obtaining interactive information: The interaction information is related to the first interaction record of multiple historical display images corresponding to the client. Optionally, the interaction information can be the first interaction record of multiple historical display images corresponding to the client, or it can be the interaction quantification value obtained by comprehensive analysis based on the first interaction record of multiple historical display images corresponding to the client. The specific settings can be flexibly configured. The aforementioned first interaction record can be obtained through the client's behavior log.

[0038] In some cases, the interaction information includes an interaction quantization value generated based on the first interaction record. The interaction quantization value is used to characterize the client's maximum tolerance for image loading time. The interaction quantization value can also be called the loading time tolerance. It is used to quantify the resistance to attenuation of interaction behavior as image loading time increases. For example, the higher the interaction quantization value, the closer the probability of interaction behavior occurring under high loading time is to the probability of interaction behavior occurring under low loading time.

[0039] In some cases, the interactive quantization value is generated through steps 1 through 4 as follows: Step 1: From the first interaction records of multiple historical display images corresponding to the client, filter out multiple second interaction records to be analyzed. In some examples, all first interaction records can be used as second interaction records. In other examples, second interaction records that meet certain conditions can be selected from the first interaction records. For example, valid second interaction records can be filtered out. First interaction records that fail to load or have abnormal loading times are excluded. For example, if the image loading time exceeds a preset normal loading time threshold, it is considered abnormal. In addition, if the first interaction record also contains auxiliary information such as the playback duration corresponding to the interaction behavior to determine whether the interaction behavior is valid, it can be used for filtering. For example, if the first interaction record contains playback duration after clicking, and the playback duration after clicking in the first interaction record is longer than the preset duration, then the first interaction record can be used as a second interaction record. This effectively excludes user accidental touches and fully ensures the reliability of the final second interaction records. In addition, if the loading time of the first interaction record is normal, and if its indication does not have a specified type of interaction behavior, such as indicating that the corresponding historical display image has not been triggered, then the first interaction record is also used as a second interaction record. In addition, in practical applications, if the total number of valid records for a user is less than a preset threshold, and the amount of data is small, the interaction quantification value for that user may not be calculated, thereby avoiding small sample error.

[0040] Step 2: Based on the image loading time of each of the multiple second interaction records, generate multiple loading time intervals. Optionally, sort the image loading times in ascending order and perform bucketing, dividing them into K buckets, each bucket corresponding to a loading time interval. This paper does not restrict the bucketing method. For example, if K=10 is used for equal-frequency bucketing, each bucket contains 10% of the second interaction records. Each loading time interval or corresponding bucket can have a corresponding sequence number. For example, arranged in ascending order of sequence number, assuming the first loading time interval (corresponding to the first bucket) corresponds to the 10% of records with the shortest loading time, the second loading time interval (corresponding to the second bucket) corresponds to the 10% of records with the second shortest loading time, and so on, with the tenth loading time interval (corresponding to the tenth bucket) corresponding to the 10th of records with the longest loading time. The boundaries of each loading time interval are determined by the actual data distribution, such as the first loading time interval corresponding to 0~0.3s, the second loading time interval corresponding to 0.3s~0.6s, etc. The above is only an example and should not be regarded as a limitation.

[0041] Step 3: Based on the second interaction records corresponding to each of the multiple loading time intervals, calculate the probability of occurrence of the interaction behavior corresponding to each of the multiple loading time intervals.

[0042] For example, for each loading time interval, the first number of second interaction records indicating the occurrence of interactive behavior corresponding to that loading time interval can be obtained, as well as the total number of second interaction records corresponding to that loading time interval. Then, the probability of interactive behavior occurring corresponding to that loading time interval is obtained based on the ratio of the first number to the second number. In this paper, interactive behavior refers to a preset type of interactive behavior, such as a valid click. Accordingly, the first number is the number of valid clicks corresponding to the loading time interval, which is also the number of second interaction records with valid click behavior. Valid click behavior can be determined by whether the playback duration after clicking is greater than a preset duration threshold. The second number is the sum of the number of valid clicks and the number of non-clicks corresponding to the loading time interval. The probability of the above-mentioned interactive behavior occurring is also known as the click-through rate. It should also be noted that if multiple loading time intervals are divided using an equal-frequency bucketing method, the number of second interaction records corresponding to each interval will be greater than zero. If other bucketing methods are used, there may be cases where the number of second interaction records corresponding to a loading time interval is zero. In such cases, the probability of interactive behavior occurring corresponding to that loading time interval can be set to zero.

[0043] Step 4: Generate interaction quantification values ​​based on the probability of interaction occurrence corresponding to each of the multiple loading time intervals. The interaction quantification values ​​obtained from steps 1 to 4 above allow for comprehensive analysis based on the divided loading time intervals and the corresponding probability of interaction occurrence, helping to ensure that the interaction quantification values ​​are more objective and accurate. For ease of understanding, in some cases, step 4 above can be performed as follows: Steps 4.1 to 4.3 Step 4.1: Based on the number of second interaction records corresponding to each of the multiple loading time intervals, obtain the weights corresponding to each of the multiple loading time intervals.

[0044] For each loading time interval, the ratio between the number of second interaction records corresponding to that interval and the total number of second interaction records to be analyzed is obtained. The weight corresponding to that loading time interval is then determined based on this ratio. If multiple loading time intervals are obtained through equal-frequency binning, the weights corresponding to each interval are theoretically the same. However, if multiple loading time intervals are obtained through non-equal-frequency binning, the weights corresponding to each interval may differ, primarily reflecting the proportion of interaction records corresponding to each bin.

[0045] Step 4.2: For each loading time interval, a weighted average is performed based on the probability and weight of the interaction behavior corresponding to the loading time interval, as well as the probability and weight of the interaction behavior corresponding to the loading time interval preceding the loading time interval, to obtain the weighted probability corresponding to the upper limit of the loading time interval; wherein, multiple loading time intervals are arranged in ascending order of image loading time.

[0046] The weighted probability mentioned above is used to characterize the probability of an interactive behavior occurring when the loading time is less than or equal to the upper limit of the loading time interval.

[0047] Step 4.3 generates an interaction quantification value based on the weighted probabilities corresponding to the upper limits of multiple loading time intervals. Through steps 4.1 to 4.3 above, the weight of each interval can be objectively measured based on the number of interaction records corresponding to each loading time interval. Based on this, a weighted cumulative processing of the probability of interaction occurrence is performed, thus obtaining the probability of interaction occurrence when the loading time is less than or equal to the upper limit of the loading time interval. For example, taking click behavior as an example, the cumulative weighted click-through rate can be accurately and objectively obtained.

[0048] In some cases, step 4.3 above can be performed with reference to steps 1) and 2) below: Step 1) Based on the weighted probabilities corresponding to the upper limits of multiple loading time intervals, determine the first loading time interval from the multiple loading time intervals. The weighted probability corresponding to the upper limit of the first loading time interval is greater than or equal to a preset probability threshold, and the weighted probability corresponding to the upper limit of the loading time interval preceding the first loading time interval is less than the loading probability value. The aforementioned probability threshold can be determined based on the overall distribution of the probability of interactive behavior occurrence, such as selecting 90% of the average probability of interactive behavior occurrence for all users, which can represent the coverage of user-preferred interactive behavior. The first loading time interval found in the above manner corresponds to the maximum loading time interval for user-preferred interactive behavior.

[0049] Step 2) Based on the ranking of the first loading time interval among multiple loading time intervals and the total number of multiple loading time intervals, normalization is performed to obtain the interaction quantization value. Specifically, the interaction quantization value can be obtained based on the ratio of the ranking to the total number. The higher the interaction quantization value, the higher the client's tolerance for loading time. Through the above steps 1) and 2), the maximum acceptable level of image loading time for the client can be objectively and reliably determined, which helps to make image file selection decisions based on this, ensure the accuracy of the file selection results, and avoid affecting the user experience due to excessive loading time.

[0050] The methods for obtaining the reference information provided in (I) to (III) above are all examples. In actual applications, any method that can obtain network status information, device performance information, and interaction information is acceptable and is not restricted here.

[0051] To facilitate understanding, this article provides a specific example of obtaining image prediction information for multiple quality levels corresponding to the first image. The image prediction information includes image loading time and image quality, which will be explained below.

[0052] Optionally, a bandwidth prediction score can be obtained, and the method of obtaining the score can refer to the aforementioned relevant content. The file size corresponding to the quality level can also be obtained. The image loading time can be determined based on the ratio of the file size to the bandwidth prediction score. For example, the product of this ratio and a preset safety factor can be used as the image loading time. The safety factor can be used to correct the download time and can be flexibly set according to network speed fluctuations. For example, when the network speed fluctuates greatly, the safety factor can be set to a larger value.

[0053] Optionally, image quality can be calculated based on one or more indicators such as structural similarity index, peak signal-to-noise ratio, and resolution. The specific quality scoring algorithm can be flexibly set. Optionally, the client can perform quality prediction, or the server can first determine the image quality of each quality level and then provide the image quality of each quality level to the client.

[0054] The above methods for obtaining image prediction information are just examples. In practical applications, any method that can obtain image loading time and image quality is acceptable and is not limited here.

[0055] In some cases, the steps in step 204 above, which involve obtaining the upper limit of loading time and the lower limit of image quality related to the client, can be performed by referring to steps A and B as follows: Step A: Obtain the interaction information corresponding to the client. This interaction information is related to the first interaction record of multiple historically displayed images on the client. The first interaction record includes at least: the image loading time corresponding to the historically displayed image and the interaction result record. The interaction result record indicates whether the historically displayed image corresponds to a preset type of interaction behavior. This article does not limit the preset type of interaction behavior. Optionally, interaction behaviors include tapping, saving, liking, forwarding, or dragging, etc. These are just examples; in actual applications, interaction behaviors can be flexibly set and are not limited here. Taking tapping as an example, the interaction result record specifically refers to whether the historically displayed image was tapped by the user. Optionally, the interaction result record can further include auxiliary information such as the playback duration corresponding to the interaction behavior to determine whether the interaction behavior is effective, in case the interaction behavior is due to user error, thereby further improving the reliability of analysis based on the first interaction record. For example, if the image mentioned in this article is a video cover image, tapping the image can directly play the corresponding video. Based on this, the first interaction record of user u for a certain historically displayed image can include the following: <image loading time> Do you want to click 'c'? Playback duration after clicking The specific methods for obtaining interactive information can be found in the aforementioned content and will not be repeated here.

[0056] Step B involves obtaining the upper limit of loading time and the lower limit of image quality related to the client, based on interaction information and device performance information. In the quality selection strategy implemented in steps A and B above, since the upper limit of loading time and the lower limit of image quality are generated based on interaction information and device performance information, they can objectively reflect the user's personalized needs and the device's processing capabilities. This helps to comprehensively analyze and select the quality level that truly matches the client's current situation, improving the accuracy of the quality selection results.

[0057] In some cases, step B above can be performed by referring to steps B1 to B2 below: Step B1: Obtain the preset upper limit benchmark value and lower limit benchmark value for time consumption. These can be flexibly set according to actual conditions. For example, relatively objective upper limit benchmark values ​​and lower limit benchmark values ​​for quality consumption can be obtained through data analysis and statistics. All clients should have the same upper limit benchmark value and lower limit benchmark value for quality consumption, and further personalized adjustments can be made based on these benchmark values ​​later.

[0058] Step B2 involves adjusting the upper limit benchmark value for loading time based on interaction information and device performance information to obtain the upper limit for loading time for the client; and adjusting the lower limit benchmark value for image quality based on device performance information to obtain the lower limit for image quality for the client. This approach fully considers individual user needs and the actual processing capabilities of the device. Compared to setting uniform upper and lower limits for all clients, steps B1 and B2 achieve the effect of customizing parameters for different clients, fully ensuring the rationality and reliability of the upper limit for loading time and the lower limit for image quality for each client.

[0059] In some cases, device performance information includes performance scores, and the lower limit of image quality corresponding to the client is positively correlated with the performance score corresponding to the client; that is, the higher the performance score, the higher the lower limit of image quality. In other words, the higher the performance score of the device corresponding to the client, the stronger the processing capability. Therefore, the lower limit of image quality can be appropriately increased to ensure the image display effect as much as possible. Interaction information includes interaction quantization values ​​generated based on the first interaction record. Interaction quantization values ​​are used to characterize the maximum acceptable level of image loading time for the client. The specific acquisition method can be obtained according to the aforementioned relevant content, and will not be repeated here. On this basis, optionally, the step of correcting the upper limit benchmark value of loading time based on interaction information and device performance information in step B2 above to obtain the upper limit of loading time corresponding to the client can be performed by referring to the following steps (1) to (3): Step (1): Adjust the baseline value of the time consumption limit based on the interaction quantization value to obtain the first adjusted upper limit; wherein, the first adjusted upper limit is positively correlated with the interaction quantization value. Optionally, a first adjustment coefficient (also called a first discount factor) can be determined based on the interaction quantization value, and the product of the first adjustment coefficient and the baseline value of the time consumption limit can be used as the first adjusted upper limit; wherein, the interaction quantization value is positively correlated with the first adjustment coefficient. In some specific examples, a sorting analysis can be performed based on the interaction quantization values ​​corresponding to multiple clients, and a first correspondence relationship between the sorting quantile interval and the first adjustment coefficient can be set, thereby determining the first adjustment coefficient based on the first correspondence relationship. Step (2) involves adjusting the baseline value of the time consumption limit based on the performance score to obtain a second adjusted upper limit; wherein the second adjusted upper limit is positively correlated with the performance score; optionally, a second adjustment coefficient (also known as a second discount factor) can be determined based on the performance score, and the product of the second adjustment coefficient and the baseline value of the time consumption limit can be used as the second adjusted upper limit; wherein the performance score is positively correlated with the second adjustment coefficient. In some specific examples, a ranking analysis can be performed based on the performance scores corresponding to multiple clients, and a second correspondence relationship between the ranking quantile interval and the second adjustment coefficient can be set, thereby determining the second adjustment coefficient based on the second correspondence relationship. Step (3): Based on the minimum of the first and second correction upper limits, obtain the upper limit of loading time for the client. That is, if the first correction upper limit is less than the second correction upper limit, the upper limit of loading time is the first correction upper limit; if the second correction upper limit is less than the first correction upper limit, the upper limit of loading time is the second correction upper limit. If the two are equal, the upper limit of loading time is equal to the two values. Based on the above steps (1) to (3), on the basis of a unified upper limit benchmark value, the upper limit benchmark value of loading time is corrected from the perspective of interactive quantitative values ​​that can present the user's personalized needs and the performance score that can present the device's processing capabilities. Then, based on the comparison of the two correction results, the smallest correction upper limit is selected, which helps to fully ensure the rationality of the final upper limit of loading time. Subsequently, the selection results can be constrained based on the upper limit of loading time, ensuring that the image loading time corresponding to the final selected quality level is less than the upper limit of loading time, thereby improving the user experience.

[0060] For easier understanding, please refer to Figure 3The diagram illustrates a process for selecting image quality levels. It shows how the upper limit of loading time can be adjusted based on interaction information and device performance information to obtain the upper limit of loading time for the client. Similarly, the lower limit of image quality can be adjusted based on device performance information to obtain the lower limit of image quality for the client. Furthermore, by combining image prediction information (image loading time and image quality) and network status information from multiple quality levels, a first quality level suitable for the client is obtained based on a multi-objective optimization strategy. The above adjustment process can be referenced in the preceding content. This method allows for personalized upper and lower limits for the client, and enables a comprehensive selection decision based on multiple dimensions. Compared to setting the same upper and lower limits for all clients and making selection decisions based on a single dimension, this approach helps to personalize decisions based on the client's current state, improving the multi-objective optimization effect.

[0061] In some cases, multi-objective optimization-based image selection strategies aim to optimize both image loading time and image quality. For example, a multi-objective optimization model can be set up. Assuming the number of first images to be displayed on the client page is L (where L is a positive integer greater than or equal to one), the optimization objective of this model is to maximize the comprehensive evaluation value of the L first images while satisfying all constraints. The comprehensive evaluation value is a value obtained based on the weighted image loading time and image quality. Constraints include at least loading time constraints and image quality constraints, and may also include image quality consistency constraints. In practical applications, to better balance the accuracy and computational efficiency of the algorithm, this paper provides a specific implementation method for the image selection strategy, which can quickly solve the multi-objective model under the above constraints. For example, it can be implemented based on a greedy algorithm, the time complexity of which is related to the total number of first quality levels, so that the algorithm can complete the decision within milliseconds or even microseconds to meet the requirements of real-time image selection. Optionally, optimal decisions can be made for each first image individually, maximizing the overall evaluation value of each first image to help maximize the overall evaluation values ​​of the L first images. Optionally, to improve selection efficiency and reduce selection difficulty, decisions can be made in stages, such as filtering based on image quality and image loading time sequentially, and subsequently, the initially determined quality levels can be corrected for consistency to meet consistency constraints. For ease of understanding, this paper provides an exemplary selection strategy. In some cases, step 206 above, selecting the first quality level suitable for the client from multiple quality levels corresponding to the first image based on at least one of network status information and device performance information, the upper limit of loading time, and the lower limit of image quality, can be performed as follows: Step a: Obtain the image quality and image loading time of the first image at multiple quality levels. That is, obtain the image prediction information of the first image at multiple quality levels. The method of obtaining this information can be referred to the aforementioned content and will not be repeated here.

[0062] Step b: Based on the image quality, image loading time, upper limit of loading time, and lower limit of image quality corresponding to each of the multiple quality levels, select the second quality level from the multiple quality levels; wherein, the image quality corresponding to the second quality level is not lower than the lower limit of image quality, and the image loading time corresponding to the second quality level is not higher than the upper limit of loading time.

[0063] Using the above method, a second quality level can be initially selected where both image quality and loading time meet the basic limits. This article does not restrict the order of time-based filtering and image quality filtering. In some examples, levels that do not meet the image quality requirements can be filtered out first, and then levels that do not meet the time requirements can be filtered out to obtain the second quality level.

[0064] Step c: Based on network status information, the image quality corresponding to the second quality level, and the image loading time, obtain the comprehensive evaluation value corresponding to the second quality level; wherein, the correlation between the comprehensive evaluation value and the image loading time, and the correlation between the comprehensive evaluation value and the image quality, are both determined based on network status information.

[0065] Optionally, network quality can be assessed based on network status information. This allows for the dynamic determination of image quality and loading time weights based on the network quality assessment. Different network qualities require different image quality and loading time weights. For example, when network quality is poor, the loading time weight is gradually increased to prioritize loading performance, while when network quality is good, the loading time weight is appropriately reduced to give more weight to image quality. Based on these image quality and loading time weights, the image quality and loading time corresponding to the second quality level can be weighted to obtain a comprehensive evaluation value for that level. This comprehensive evaluation value characterizes the compatibility between the second quality level and the client.

[0066] Step d: Select the third quality level from the second quality level, and determine the first quality level that is compatible with the client based on the third quality level; wherein, the third quality level is the second quality level with the highest comprehensive evaluation value.

[0067] Optionally, the third quality level can be directly used as the first quality level adapted to the client, or further analysis and processing can be performed based on the third quality level, such as adjusting it if the third quality level does not meet other constraints, so as to ensure that the final first quality level adapted to the client can meet all constraints and achieve optimal loading time and image quality as much as possible.

[0068] By using steps a to d above, we can first filter out the second quality level, which meets the basic limit requirements for both image quality and loading time. Based on this, we can then combine network status information, the image quality and image loading time corresponding to the second quality level to comprehensively evaluate the second quality level and select the second quality level with the highest comprehensive evaluation value. This method of determining the first quality level that is compatible with the client is more accurate and reliable. Moreover, the above step-by-step selection method is simpler and easier to implement, without the need for tedious and complex analysis and calculations, and has higher selection efficiency.

[0069] In some cases, network status information includes network quality values, and step c above can be performed by referring to steps c1 and c2 below: Step c1: Based on the comparison result between the network quality value and the preset network quality lower limit, determine the first weight corresponding to the image quality and the second weight corresponding to the image loading time; wherein, if the comparison result indicates that the network quality value is not higher than the preset network quality lower limit, the first weight is the preset value; if the comparison result indicates that the network quality value is higher than the preset network quality lower limit, the first weight is positively correlated with the network quality value; the sum of the first weight and the second weight is one.

[0070] The network quality lower limit can be an adjustable threshold used to distinguish between weak and strong networks. In practical applications, the above preset value can be flexibly set; optionally, the preset value is zero. That is, if the network quality value is less than or equal to the preset network quality lower limit, it indicates that the current network is poor. In this case, the first weight of image quality is set to zero, and the corresponding second weight of image loading time is set to one. In other words, image quality is no longer considered in the selection process; only image loading time is considered. If the network quality value is higher than the preset network quality lower limit, it indicates that the current network is acceptable. In this case, image quality can be considered, and the higher the network quality value, the greater the first weight of image quality, and the smaller the corresponding second weight of image loading time. Optionally, when the network quality value is higher than the preset network quality lower limit, the first weight is equal to the ratio of the first difference to the second difference; where the first difference is equal to the difference between the network quality value and the network quality lower limit, and the second difference is equal to the difference between the preset network quality upper limit and the network quality lower limit.

[0071] Step c2: Based on the first weight, the second weight, and the image quality and image loading time of the second quality level, obtain the comprehensive evaluation value corresponding to the second quality level.

[0072] Optionally, the comprehensive evaluation value is equal to the sum of the first product value and the second product value, wherein the first product value is equal to the product of the first weight and the image quality, the second product value is equal to the difference between the first weight and the time consumption ratio, and the time consumption ratio is equal to the ratio between the image loading time of the second quality level and the upper limit of the loading time. Through steps c1 to c2 above, a comprehensive evaluation of the second quality level can be reasonably performed based on network status information, image quality corresponding to the second quality level, and image loading time. The comprehensive evaluation value is negatively correlated with image loading time. When the network quality is poor, the comprehensive evaluation value is only related to image loading time. When the network quality is good, the comprehensive evaluation value is positively correlated with image quality and negatively correlated with image loading time. That is, when solving the first quality level for multi-objective optimization and client adaptation, the weight of network status on image quality and image loading time can be fully considered. For example, when the network quality is poor, the weight of loading time is gradually increased to prioritize loading performance, while when the network quality is good, the weight of loading time is appropriately reduced to give more weight to quality. This effectively ensures the rationality and reliability of the final comprehensive evaluation value and further helps to improve the accuracy of the quality level selection result.

[0073] In some cases, there are multiple images in the first image. To avoid the client displaying some images in high definition and others in blurry image quality on the same page, this paper sets an additional image quality consistency constraint. That is, the standard deviation of image quality between images on the same page should be less than a preset standard deviation threshold to ensure that the overall image quality of all images on the same page is consistent, thereby providing users with a better visual experience. Based on this, the step in step d above, which determines the first quality level adapted to the client based on the third quality level, can be performed as follows: steps d1 to d3: Step d1: Based on the image quality of each of the multiple first images corresponding to the third quality level, obtain the first quality standard deviation corresponding to the client page; wherein, the client page is a page that displays multiple first images simultaneously; the first quality standard deviation can be calculated using the standard deviation formula, which will not be elaborated here.

[0074] Step d2: In response to the first quality standard deviation not exceeding a preset standard deviation threshold, the third quality level corresponding to the first image is taken as the first quality level adapted to the client. Optionally, the standard deviation threshold can be determined through user experience testing. If the first quality standard deviation is less than or equal to the standard deviation threshold, it means that the third quality level of all first images on the same client page meets the overall quality consistency constraint. Therefore, the third quality level of each first image (i.e., the optimal second quality level) can be directly taken as the first quality level adapted to the client.

[0075] Step d3: In response to the first quality standard deviation being higher than the standard deviation threshold, a quality level correction operation is performed to select the first quality level adapted to the client from the second quality levels corresponding to the first image. The second quality standard deviation corresponding to the client page is not higher than the standard deviation threshold, and is obtained based on the image quality of the first quality levels corresponding to multiple first images. If the first quality standard deviation is higher than the standard deviation threshold, it indicates that the third quality level of all first images on the same client page does not meet the quality consistency constraint. In this case, the quality levels of at least some first images need to be corrected by replacing the third quality level with other second quality levels, so that the final first quality levels adapted to the client for all first images meet the quality consistency constraint. For each first image to be displayed on the client page, the first quality level adapted to the client for that first image is obtained in the above manner.

[0076] Through the above steps d1 to d3, the image quality corresponding to the selected third quality level (i.e. the optimal second quality level) can be used as the measurement object. Based on the quality standard deviation, the image quality presentation effect between different images on the same page is objectively measured to see if there is a large difference. If it does not meet the requirements, it is corrected. In this way, while taking into account both image quality and loading time, the overall visual experience of the user can be effectively guaranteed.

[0077] In some cases, step d3 above involves performing a quality level correction operation, including: performing at least one quality level correction operation with the objective of minimizing the total image quality loss corresponding to multiple first images; wherein the total image quality loss is determined based on the sum of the image quality losses corresponding to each of the multiple first images, the image quality loss corresponding to the first image is the difference between the image quality at the current quality level and the average quality, the current quality level is the quality level currently corresponding to the quality level correction operation, and the average quality is the average of the image quality at the current quality level corresponding to the multiple first images. The quality level currently corresponding to the first quality level correction operation is the third quality level, and the quality level currently corresponding to subsequent quality level correction operations is the quality level selected in the previous quality level correction operation.

[0078] By using the above-mentioned level correction method, the overall image quality loss can be minimized, and the impact on quality can be reduced as much as possible while meeting the quality consistency constraint.

[0079] Optionally, to efficiently achieve the above objectives, when performing the quality level correction operation, the median quality of the third quality level corresponding to multiple first images on the client page can be determined first. The upper limit of the standard deviation is determined based on the sum of the quality median and the standard deviation threshold, and the lower limit of the standard deviation is determined based on the difference between the quality median and the standard deviation threshold. For each first image, the image quality of the current quality level corresponding to that first image is obtained. If the image quality of the current quality level corresponding to that first image is greater than the upper limit of the standard deviation, the fourth quality level of the first image is selected from the second quality levels corresponding to that first image. This fourth quality level is taken as the currently corrected quality level of the first image. Then, the first quality level adapted to the client can be further determined based on the fourth quality level. Here, the fourth quality level is the quality level among the second quality levels corresponding to the first image where the image quality is higher than the quality median and closest to the quality median. In response to the image quality of the current quality level corresponding to the first image being less than the lower limit of the standard deviation, the fifth quality level corresponding to the first image is selected from the second quality levels corresponding to the first image. This fifth quality level is then used as the current corrected quality level for the first image. Subsequently, a first quality level adapted to the client can be determined based on this fifth quality level. The fifth quality level is the level among the second quality levels corresponding to the first image where the image quality is lower than the median but closest to it. The median is used in this method because it is more resistant to extreme values ​​than the mean, allowing for more reliable quality level correction. The sum of the median and the standard deviation threshold effectively ensures the reasonableness of the upper limit of the standard deviation. If the image quality of the current quality level is greater than the upper limit of the standard deviation, it indicates that the quality of the current quality level is too high, thus requiring a reduction in quality. The preferred quality level is one where the image quality is higher than the median and closest to it, thus helping to ensure that the corrected quality level is closest to the median while still maintaining a high quality effect. Correspondingly, the difference between the median quality and the standard deviation threshold can effectively ensure the reasonableness of the lower limit of the standard deviation. If the image quality of the current quality level is less than the lower limit of the standard deviation, it indicates that the quality of the current quality level is too low, and therefore the quality needs to be improved. The preferred quality level is one whose image quality is lower than the median quality but closest to it. This helps to ensure that the quality of the corrected quality level is closest to the median quality while still maintaining a relatively low quality effect. If the result of a single quality level correction operation still does not meet the one-time quality constraint, the above operation can be repeated multiple times until the one-time quality constraint is met or until a preset number of operations is reached. This method can efficiently achieve quality consistency correction processing and effectively avoid problems such as excessive quality loss and excessive computational overhead caused by over-correction.

[0080] See Figure 4 The diagram illustrates a process for selecting image quality levels. Figure 3 Building upon this, a specific implementation example of a multi-objective optimization-based quality selection strategy is further illustrated. Available quality levels can be filtered based on image quality constraints and loading time constraints to obtain a second quality level. Then, a comprehensive evaluation of the second quality level is performed, and the quality level with the highest comprehensive evaluation score is selected as the third quality level. Subsequently, a quality consistency correction process based on minimizing loss can be used to obtain a first quality level adapted to the client. The specific implementation method can be found in the aforementioned related content and will not be repeated here. Through this method, the final selected quality level can be optimized as much as possible while meeting constraints such as loading time and quality, thereby improving the user experience.

[0081] For ease of understanding, in practical applications, please also refer to... Figure 5 The diagram illustrates a process for selecting image quality levels, demonstrating the client-side strategy for image quality level selection. This involves three key steps: input signal acquisition and calculation (including network status information, device performance information, and interaction information), personalized parameter decision-making (lower limit for image quality, upper limit for loading time), and image quality level selection. It can be understood that the same source image undergoes multiple transcodings with different parameters, resulting in multiple images with the same content but different resolutions, image qualities, and data sizes. These images constitute a set of multi-bitrate products of the source image. For example, an original 2K resolution image, after transcoding at multiple quality levels, yields multiple versions such as 1080p, 720p, and 540p. This set of image versions is the multi-bitrate product of the source image, with each transcoded image corresponding to a quality level. The image prediction information for the aforementioned quality level corresponds to the image information of each transcoded image version. Specifically, the server can provide the client with image information corresponding to multiple quality levels, i.e., the aforementioned image description information. This includes a set of structured data describing key information of the multi-bitrate transcoding products of the source image, such as the resolution, data size, and storage address of each transcoded image on the server. There is a one-to-one correspondence between the source image and the first image to be displayed on the client; the content of the source image is the content of the first image to be displayed on the client. Figure 5 In this diagram, different colors represent different source images, and the same color represents different transcoding results of the same source image. After staged selection and quality correction, the final selection result is obtained, which is the first quality level adapted to the client for each first image. The specific selection method can be found in the aforementioned content and will not be repeated here.

[0082] In summary, the image quality level selection method presented in this paper can integrate multi-dimensional information such as network status, device performance, and interaction information in complex application environments. Based on this, the image quality selection strategy helps to maximize image display quality while ensuring loading performance, effectively solving multi-objective optimization problems such as loading time and image quality. Furthermore, the strategy fully considers the different needs of different users, providing customized parameters such as lower limits for image quality and upper limits for loading time. In conclusion, the image quality selection strategy based on multi-dimensional information perception and multi-objective optimization not only ensures image quality and accurately conveys content information (e.g., under image quality constraints, short video covers have scene recognition, and text / image covers have theme prominence), but also considers loading efficiency under different network environments and adaptability to different devices. It also fully considers the differences in user needs for speed and image quality, and the selection results can effectively adapt to users' personalized needs. Ultimately, it helps to achieve a globally optimal balance between image loading performance (corresponding to loading time) and display quality, making it well-suited for media resource platforms and contributing to improved information delivery efficiency and optimized user experience.

[0083] Corresponding to the aforementioned method for selecting image quality levels, this paper further provides a device for selecting image quality levels. Figure 6 This paper presents a schematic diagram of an image quality level selection device. This device can be implemented by software and / or hardware, and is generally integrated into electronic devices, such as... Figure 6 As shown, the image quality level selection device includes: The request receiving module 602 is used to obtain a first request received by the client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; The information acquisition module 604 is used to acquire at least one of the network status information and device performance information of the electronic device where the client is located, and to acquire the upper limit of loading time and the lower limit of image quality related to the client. The quality selection module 606 is used to select a first quality level that is compatible with the client from multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit of loading time and the lower limit of image quality. The image display module 608 is used to display the first image on the client according to the selected first quality level.

[0084] The aforementioned device fully considers that the accuracy of selecting quality levels based solely on a single factor in related technologies is poor. Since image loading time and image quality corresponding to different quality levels are the main factors affecting user experience, it selects a quality level that is compatible with the client based on multiple reference factors, such as network status information, device performance information, upper limit of loading time related to the client, and lower limit of image quality. This can effectively improve the accuracy of image selection results and help ensure that the first image displayed by the client can better meet the image quality and loading time requirements, thereby improving the user experience.

[0085] In some cases, the information acquisition module 604 is specifically used to: acquire the interaction information corresponding to the client; the interaction information is information related to a first interaction record of multiple historical display images corresponding to the client; the first interaction record includes at least: the image loading time and interaction result record corresponding to the historical display image, the interaction result record being used to indicate whether the historical display image corresponds to a preset type of interaction behavior; based on the interaction information and the device performance information, acquire the upper limit of loading time and the lower limit of image quality related to the client.

[0086] In some cases, the information acquisition module 604 is specifically used to: acquire a preset upper limit benchmark value for loading time and a lower limit benchmark value for quality; correct the upper limit benchmark value for loading time based on the interaction information and the device performance information to obtain the upper limit for loading time corresponding to the client; and correct the lower limit benchmark value for loading time based on the device performance information to obtain the lower limit for image quality corresponding to the client.

[0087] In some cases, the device performance information includes a performance score, and the lower limit of image quality corresponding to the client is positively correlated with the performance score corresponding to the client; the interaction information includes an interaction quantization value generated based on the first interaction record, and the interaction quantization value is used to characterize the maximum acceptable level of the client to image loading time.

[0088] In some cases, the information acquisition module 604 is specifically used to: correct the time consumption upper limit benchmark value based on the interaction quantization value to obtain a first corrected upper limit; wherein the first corrected upper limit is positively correlated with the interaction quantization value; correct the time consumption upper limit benchmark value based on the performance score to obtain a second corrected upper limit; wherein the second corrected upper limit is positively correlated with the performance score; and obtain the loading time upper limit corresponding to the client based on the minimum value of the first corrected upper limit and the second corrected upper limit.

[0089] In some cases, the device further includes an interaction quantization value generation module, used to generate the interaction quantization value through the following steps: filtering out multiple second interaction records to be analyzed from a first interaction record of multiple historical display images corresponding to the client; generating multiple loading time intervals based on the image loading time of each of the multiple second interaction records; calculating the probability of occurrence of the interaction behavior corresponding to each of the multiple loading time intervals based on the second interaction records corresponding to each of the multiple loading time intervals; and generating an interaction quantization value based on the probability of occurrence of the interaction behavior corresponding to each of the multiple loading time intervals.

[0090] In some cases, the interaction quantization value generation module is specifically used for: generating interaction quantization values ​​based on the probability of occurrence of interactive behaviors corresponding to each of the multiple loading time intervals; obtaining the weights corresponding to each of the multiple loading time intervals based on the number of second interaction records corresponding to each of the multiple loading time intervals; for each loading time interval, performing weighted processing based on the probability and weight of occurrence of interactive behaviors corresponding to the loading time interval, and the probability and weight of occurrence of interactive behaviors corresponding to the loading time interval preceding the loading time interval, to obtain the weighted probability corresponding to the upper limit of the loading time interval; wherein, the multiple loading time intervals are arranged in ascending order of image loading time; and generating interaction quantization values ​​based on the weighted probabilities corresponding to the upper limits of the multiple loading time intervals.

[0091] In some cases, the interaction quantization value generation module is specifically used to: determine a first time interval from the multiple loading time intervals based on the weighted probabilities corresponding to the upper limits of the loading time intervals; wherein the weighted probability corresponding to the upper limit of the first time interval is greater than or equal to a preset probability threshold, and the weighted probability corresponding to the upper limit of the loading time interval preceding the first time interval is less than the loading probability value; and perform normalization processing based on the ranking of the first time interval among the multiple loading time intervals and the total number of the multiple loading time intervals to obtain the interaction quantization value.

[0092] In some cases, the quality level selection module 606 is specifically used for: obtaining the image quality and image loading time corresponding to the first image at multiple quality levels; selecting a second quality level from the multiple quality levels based on the image quality and image loading time corresponding to each of the multiple quality levels, the upper limit of loading time, and the lower limit of image quality; wherein the image quality corresponding to the second quality level is not lower than the lower limit of image quality, and the image loading time corresponding to the second quality level is not higher than the upper limit of loading time; obtaining a comprehensive evaluation value corresponding to the second quality level based on the network status information, the image quality and image loading time corresponding to the second quality level; wherein the correlation between the comprehensive evaluation value and the image loading time, and the correlation between the comprehensive evaluation value and the image quality, are both determined based on the network status information; selecting a third quality level from the second quality levels, and determining a first quality level adapted to the client based on the third quality level; wherein the third quality level is the second quality level with the highest comprehensive evaluation value.

[0093] In some cases, the grade selection module 606 is specifically used to: determine a first weight corresponding to the image quality and a second weight corresponding to the image loading time based on the comparison result between the network quality value and a preset network quality lower limit; wherein, if the comparison result indicates that the network quality value is not higher than the preset network quality lower limit, then the first weight is a preset value; if the comparison result indicates that the network quality value is higher than the preset network quality lower limit, then the first weight is positively correlated with the network quality value; the sum of the first weight and the second weight is one; and based on the first weight, the second weight, and the image quality and image loading time of the second quality grade, obtain a comprehensive evaluation value corresponding to the second quality grade.

[0094] In some cases, the number of first images is multiple, and the quality level selection module 606 is specifically used to: obtain a first quality standard deviation corresponding to the client page based on the image quality of the third quality level corresponding to each of the multiple first images; wherein the client page is a page that displays multiple first images simultaneously; in response to the first quality standard deviation not being higher than a preset standard deviation threshold, using the third quality level corresponding to the first image as the first quality level adapted to the client; in response to the first quality standard deviation being higher than the standard deviation threshold, performing a quality level correction operation to select the first quality level adapted to the client from the second quality levels corresponding to the first image, wherein the second quality standard deviation corresponding to the client page is not higher than the standard deviation threshold, and the second quality standard deviation is obtained based on the image quality of the first quality level corresponding to each of the multiple first images.

[0095] In some cases, the gear selection module 606 is specifically used to: perform at least one gear correction operation with the objective of minimizing the total image quality loss corresponding to multiple first images; wherein the total image quality loss is determined based on the sum of the image quality losses corresponding to each of the multiple first images, the image quality loss corresponding to the first image is the difference between the image quality of the current quality gear corresponding to the first image and the average quality, the current quality gear is the quality gear currently corresponding to the gear correction operation, and the average quality is the average value of the image quality of the current quality gear corresponding to multiple first images.

[0096] The image quality level selection device provided in this paper can execute the image quality level selection method provided in any example of this paper, and has the corresponding functional modules and beneficial effects of the execution method.

[0097] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the device example described above can be referred to the corresponding process in the method example, and will not be repeated here.

[0098] This article provides an electronic device, which includes: a storage device storing a computer program thereon; and a processing device for executing the computer program in the storage device to implement the steps of any of the image quality levels selection methods described herein. The beneficial effects that can be achieved include, but are not limited to, effectively improving the accuracy of image quality selection results and enhancing the user experience.

[0099] The following is for reference. Figure 7 This document illustrates a schematic diagram of the structure of an electronic device 700 suitable for implementing the present invention. The terminal devices described herein may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Devices), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital televisions and desktop computers. Figure 7 The electronic device shown is merely an example and should not impose any limitations on the functionality and scope of this article.

[0100] like Figure 7As shown, the electronic device 700 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device 700. The processing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0101] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic device 700 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 700 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0102] In particular, according to the examples herein, the processes described in the above-referenced flowcharts can be implemented as computer software programs. For example, the examples herein include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such an example, the computer program can be downloaded and installed from a network via communication device 709, or installed from storage device 708, or installed from ROM 702. When the computer program is executed by processing device 701, it performs the functions defined in the methods herein.

[0103] Furthermore, the examples in this article can also be computer-readable storage media storing a computer program that, when executed by a processor, implements the image quality level selection method described in any of the claims herein, and the beneficial effects that can be achieved include, but are not limited to, effectively improving the accuracy of image quality level selection results and enhancing the user experience.

[0104] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0105] In addition to the methods, devices, and media described above, examples herein can also be computer program products comprising a computer program that, when executed by a processor, implements the image quality level selection method described in any of the claims provided herein. Exemplarily, the computer program product is stored in a computer storage medium and includes computer-executable instructions that, when executed by a device, cause the device to perform the image quality level selection method according to any of the claims provided herein. The resulting benefits include, but are not limited to, effectively improving the accuracy of image quality level selection results and enhancing the user experience.

[0106] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations described herein. These programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0107] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0108] The above descriptions are merely specific embodiments of this document, enabling those skilled in the art to understand or implement it. Various modifications to these examples will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other examples without departing from the spirit or scope of this document. Therefore, this document is not to be limited to the examples described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for selecting an image quality level, comprising: Obtain the first request received by the client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; Obtain at least one of the network status information and device performance information of the electronic device where the client is located, and obtain the upper limit of loading time and the lower limit of image quality related to the client; Based on at least one of the network status information and device performance information, the upper limit of loading time and the lower limit of image quality, a first quality level that is compatible with the client is selected from multiple quality levels corresponding to the first image. The first image is displayed on the client according to the selected first quality level.

2. The method according to claim 1, wherein obtaining the upper limit of loading time and the lower limit of quality related to the client includes: Obtain the interaction information corresponding to the client; The interaction information is information related to the first interaction record of multiple historical display images corresponding to the client; The first interaction record includes at least: the image loading time and interaction result record corresponding to the historical display image, wherein the interaction result record is used to indicate whether the historical display image corresponds to a preset type of interaction behavior; Based on the interaction information and the device performance information, the upper limit of loading time and the lower limit of image quality related to the client are obtained.

3. The method according to claim 2, wherein obtaining the upper limit of loading time and the lower limit of image quality related to the client based on the interaction information and the device performance information includes: Obtain the preset upper limit benchmark value for time consumption and the lower limit benchmark value for quality; Based on the interaction information and the device performance information, the baseline value of the upper limit of the time consumption is corrected to obtain the upper limit of the loading time corresponding to the client. Furthermore, the lower quality limit benchmark value is corrected based on the device performance information to obtain the lower image quality limit corresponding to the client.

4. The method according to claim 3, wherein the device performance information includes a performance score, and the lower limit of image quality corresponding to the client is positively correlated with the performance score corresponding to the client; The interaction information includes an interaction quantization value generated based on the first interaction record, which is used to characterize the client's maximum tolerance for image loading time.

5. The method according to claim 4, wherein the step of correcting the upper limit benchmark value of the time consumption based on the interaction information and the device performance information to obtain the upper limit of the loading time corresponding to the client includes: The time consumption upper limit baseline value is corrected based on the interaction quantization value to obtain a first correction upper limit; wherein, the first correction upper limit is positively correlated with the interaction quantization value; The baseline value of the time consumption limit is adjusted based on the performance score to obtain a second adjusted upper limit; wherein, the second adjusted upper limit is positively correlated with the performance score; The loading time limit for the client is obtained based on the minimum value between the first and second correction limits.

6. The method according to claim 4, wherein the interactive quantization value is generated through the following steps: From the first interaction records of multiple historical display images corresponding to the client, multiple second interaction records to be analyzed are selected; Based on the image loading time of each of the multiple second interaction records, multiple loading time intervals are generated; Based on the second interaction records corresponding to each of the multiple loading time intervals, the probability of occurrence of the interaction behavior corresponding to each of the multiple loading time intervals is calculated. Based on the probability of interactive behavior occurring in each of the multiple loading time intervals, an interaction quantification value is generated.

7. The method according to claim 6, generating an interaction quantification value based on the probability of interaction occurrence corresponding to each of the plurality of loading time intervals; Based on the number of second interaction records corresponding to each of the multiple loading time intervals, obtain the weight corresponding to each of the multiple loading time intervals; For each loading time interval, a weighted average is performed based on the probability and weight of the interaction behavior corresponding to that loading time interval, and the probability and weight of the interaction behavior corresponding to the loading time intervals preceding that loading time interval, to obtain the weighted probability corresponding to the upper limit of the loading time interval; wherein, The multiple loading time intervals are arranged in ascending order of image loading time; An interaction quantization value is generated based on the weighted probability corresponding to the upper limit of the loading time intervals for each of the multiple loading time intervals.

8. The method according to any one of claims 1 to 7, wherein selecting a first quality level adapted to the client from a plurality of quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit of loading time, and the lower limit of image quality includes: Obtain the image quality and image loading time of the first image at multiple quality levels; Based on the image quality and image loading time corresponding to each of the multiple quality levels, the upper limit of loading time, and the lower limit of image quality, a second quality level is selected from the multiple quality levels; wherein, the image quality corresponding to the second quality level is not lower than the lower limit of image quality, and the image loading time corresponding to the second quality level is not higher than the upper limit of loading time. Based on the network status information, the image quality corresponding to the second quality level, and the image loading time, a comprehensive evaluation value corresponding to the second quality level is obtained; wherein, the correlation between the comprehensive evaluation value and the image loading time, and the correlation between the comprehensive evaluation value and the image quality, are both determined based on the network status information; A third quality level is selected from the second quality level, and a first quality level adapted to the client is determined based on the third quality level; wherein the third quality level is the second quality level with the highest comprehensive evaluation value.

9. The method according to claim 8, wherein the network status information includes a network quality value, and the step of obtaining a comprehensive evaluation value corresponding to the second quality level based on the network status information, the image quality corresponding to the second quality level, and the image loading time includes: Based on the comparison result between the network quality value and the preset network quality lower limit, a first weight corresponding to the image quality and a second weight corresponding to the image loading time are determined; wherein, if the comparison result indicates that the network quality value is not higher than the preset network quality lower limit, then the first weight is a preset value; if the comparison result indicates that the network quality value is higher than the preset network quality lower limit, then the first weight is positively correlated with the network quality value; the sum of the first weight and the second weight is one. Based on the first weight, the second weight, and the image quality and image loading time of the second quality level, a comprehensive evaluation value corresponding to the second quality level is obtained.

10. The method according to claim 8, wherein the number of the first images is multiple, and the step of determining the first quality level adapted to the client based on the third quality level includes: Based on the image quality of each of the first images corresponding to the third quality level, the first quality standard deviation corresponding to the client page is obtained; wherein, the client page is a page that displays multiple first images simultaneously; In response to the first quality standard deviation not being higher than a preset standard deviation threshold, the third quality level corresponding to the first image is used as the first quality level adapted to the client. In response to the first quality standard deviation being higher than the standard deviation threshold, a quality level correction operation is performed to select a first quality level that is compatible with the client from the second quality levels corresponding to the first image, wherein the second quality standard deviation corresponding to the client page is not higher than the standard deviation threshold, and the second quality standard deviation is obtained based on the image quality of the first quality level corresponding to each of the multiple first images.

11. The method according to claim 10, wherein performing the gear shift correction operation comprises: With the objective of minimizing the total image quality loss corresponding to multiple first images, at least one quality level correction operation is performed; wherein, the total image quality loss is determined based on the sum of the image quality losses corresponding to each of the multiple first images, the image quality loss corresponding to the first image is the difference between the image quality at the current quality level corresponding to the first image and the average quality, the current quality level is the quality level currently corresponding to the quality level correction operation, and the average quality is the average of the image quality at the current quality level corresponding to multiple first images.

12. A device for selecting image quality levels, comprising: The request receiving module is used to obtain a first request received by the client, wherein the first request includes a request to display a first image, and the first image corresponds to multiple quality levels; The information acquisition module is used to acquire at least one of the network status information and device performance information of the electronic device where the client is located, as well as to acquire the upper limit of loading time and the lower limit of image quality related to the client; The quality selection module is used to select a first quality level that is compatible with the client from multiple quality levels corresponding to the first image based on at least one of the network status information and device performance information, the upper limit of loading time and the lower limit of image quality. An image display module is used to display the first image on the client according to the selected first quality level.

13. An electronic device, the electronic device comprising: A storage device on which computer programs are stored; A processing device for executing the computer program in the storage device to implement the steps of the method for selecting an image quality level according to any one of claims 1-11.

14. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for selecting an image quality level according to any one of claims 1-11.

15. A computer program product stored in a computer storage medium and comprising computer-executable instructions that, when executed by a device, cause the device to perform the method according to any one of claims 1 to 11.

Images