Image rendering methods and apparatus, electronic equipment

By analyzing the complexity of virtual digital human behavior and terminal performance, dynamically calculating the stress score of the rendering area, and adjusting the model accuracy and frame rate, the problems of high rendering computation and poor scheduling of complex behaviors are solved, achieving precise allocation of rendering resources and improved user experience.

CN123123641APending Publication Date: 2026-11-17MIGU COMIC CO LTD +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610982350.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing technologies, virtual digital human rendering methods result in a large amount of rendering computation, which can easily lead to stuttering and latency, make it difficult to be compatible with terminal devices of different performance levels, and have poor performance in scheduling complex behaviors.

Method used

By acquiring the behavioral complexity and terminal performance score of the real-time actions of digital humans, the stress score of different rendering areas is calculated, and the model accuracy and rendering frame rate are adjusted according to the stress score to achieve precise allocation of rendering resources.

Benefits of technology

It improves the smoothness and visual consistency of digital human presentation, enhances adaptability to terminals with different performance and network environments, and solves the problem of uneven allocation of rendering resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123641A_ABST
    Figure CN123123641A_ABST
Patent Text Reader

Abstract

The image rendering method, apparatus, and electronic device disclosed herein relate to the field of image rendering technology. By comprehensively analyzing the behavioral complexity of real-time actions of digital humans, the performance score of terminal devices, and the preset priorities of different rendering regions, the stress score of each rendering region is dynamically calculated to adjust the model accuracy. Simultaneously, the rendering frame rate of each rendering region is adjusted according to the behavioral complexity and terminal performance, so that the allocation of rendering resources can match actual needs and terminal capabilities. It can effectively release the rendering load of non-critical regions while ensuring high-quality rendering of critical regions. It solves the problems of high computing power consumption of full-frame and full-region rendering, poor experience in weak network environments, and poor performance of local rendering methods in scheduling complex behaviors. Therefore, this application can achieve precise allocation of rendering resources and improve the presentation quality of digital humans and user experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This disclosure relates to the field of image rendering technology, and in particular to an image rendering method, apparatus, and electronic device. Background Technology

[0002] Currently, virtual digital humans are widely used in scenarios such as live streaming, interactive all-in-one machines, Q&A screens, and metaverse. The rendering method of virtual digital humans usually adopts a full-frame, full-area rendering strategy to process each frame of animation. The rendering computation is large and highly dependent on the computing power and network bandwidth of the terminal device. Therefore, it is easy to have problems such as stuttering and latency, which affects the user experience.

[0003] To reduce rendering overhead, related technologies employ either voice-driven methods to render specific regions of the digital human independently, or use fixed strategies to reduce overall rendering quality. However, these rendering methods are ineffective at managing the complex behaviors of digital humans and struggle to achieve compatibility with terminal devices of varying performance levels.

[0004] Therefore, how to achieve precise allocation of rendering resources to improve the presentation quality of digital humans and user experience is an urgent problem to be solved. Summary of the Invention

[0005] This disclosure provides an image rendering method, apparatus, and electronic device. Its main purpose is to solve the problem of how to achieve precise allocation of rendering resources in order to improve the presentation quality of digital humans and user experience.

[0006] According to a first aspect of this disclosure, an image rendering method is provided, comprising:

[0007] Get the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, and get the terminal performance score corresponding to the terminal device. The terminal device is used to render the digital human. Based on behavioral complexity, terminal performance score, and preset region priority, calculate the stress score corresponding to each rendering region in the digital human, and adjust the model accuracy of each rendering region according to the stress score. Adjust the rendering frame rate of different rendering regions based on behavior complexity and terminal performance score; Based on the adjusted model accuracy, different rendering regions are rendered according to their respective adjusted rendering frame rates to obtain the rendering results of each rendering region. The rendering results are then merged to obtain the rendered video stream.

[0008] According to a second aspect of this disclosure, an image rendering apparatus is provided, comprising: The acquisition unit is used to acquire the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, and to acquire the terminal performance score corresponding to the terminal device, which is used to render the digital human. The calculation unit is used to calculate the stress score corresponding to different rendering areas in the digital human based on behavioral complexity, terminal performance score and preset area priority. The first adjustment unit is used to adjust the model accuracy of different rendering areas according to the pressure score; The second adjustment unit is used to adjust the rendering frame rate of different rendering regions according to the behavior complexity and terminal performance score. The rendering unit is used to render different rendering regions according to their respective adjusted rendering frame rates based on the adjusted model accuracy, so as to obtain the rendering results of different rendering regions. The merging unit is used to merge the rendering results to obtain the rendered video stream.

[0009] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory that is communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor is able to perform the method described in the first aspect above.

[0010] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method of the first aspect described above.

[0011] According to a fifth aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method as described in the first aspect above.

[0012] The image rendering method, apparatus, and electronic device disclosed herein, through comprehensive analysis of the behavioral complexity of the real-time actions of the digital human, the performance score of the terminal device, and the preset priorities of different rendering regions, dynamically calculates the stress score of each rendering region to adjust the model accuracy, and simultaneously adjusts the rendering frame rate of each rendering region according to the behavioral complexity and terminal performance. This allows the allocation of rendering resources to match actual needs and terminal capabilities, effectively releasing the rendering load of non-critical regions while ensuring high-quality rendering of critical areas. It solves the problems of high computational consumption of full-frame and full-region rendering, poor experience in weak network environments, and poor performance of local rendering methods in scheduling complex behaviors. Therefore, this application can improve the fluency and visual consistency of the digital human, enhance its adaptability to terminals with different performance and network environments, achieve precise allocation of rendering resources, and improve the presentation quality and user experience of the digital human.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0014] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 This is a schematic flowchart of an image rendering method provided in an embodiment of the present disclosure; Figure 2 This is a schematic diagram of the structure of an image rendering apparatus provided in an embodiment of the present disclosure; Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of the present disclosure. Detailed Implementation

[0015] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0016] This application provides an image rendering method that can be applied to scenarios requiring real-time rendering of virtual digital humans, such as digital human live streaming, digital human customer service, digital human interactive all-in-one machines, and metaverse communities. The executing entity of this method can be a terminal device with computing capabilities, such as a smartphone, tablet, personal computer, or interactive large screen, or a cloud server. When the method runs on a cloud server, the cloud server completes the rendering and pushes the rendering result to the terminal device as a video stream for display, i.e., a cloud rendering mode is adopted; when the method runs directly on the terminal device, the terminal device completes the rendering and display locally, i.e., a terminal-side rendering mode is adopted. This invention does not impose any limitations on this.

[0017] The image rendering method, apparatus, and electronic device of this disclosure are described below with reference to the accompanying drawings.

[0018] Figure 1 This is a schematic flowchart of an image rendering method provided in an embodiment of the present disclosure.

[0019] like Figure 1 As shown, the method includes the following steps: Step 101: Obtain the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, and obtain the terminal performance score corresponding to the terminal device. The terminal device is used to render the digital human.

[0020] In the embodiments of this application, the video stream to be rendered can be a pre-recorded sequence of digital human animations, or a digital human broadcast or interactive screen generated and streamed in real time. The digital human will perform real-time actions in the video stream, such as blinking, nodding, waving, turning around, walking, etc.

[0021] In this application, behavioral complexity is a comprehensive parameter used to characterize the fineness, amplitude, and coordination of multiple parts of a digital human's actions at the current moment or in the current frame. In one instance, if the digital human only performs a simple blinking action, its behavioral complexity is low; while when the digital human simultaneously performs multiple actions such as waving its hands, rotating its head significantly, and twisting its torso, its behavioral complexity is high.

[0022] Simultaneously, it is also necessary to obtain the terminal performance score of the terminal device used to render the digital human. The terminal performance score is a numerical value reflecting the current graphics rendering capability of the terminal device. The higher the score, the stronger the computing power of the terminal device and the more precise the rendering it can support. The terminal performance score can be calculated in real time based on the hardware parameters of the terminal device, including but not limited to the CPU's clock speed or utilization rate, the GPU's floating-point arithmetic capability, remaining memory capacity, current battery level, and temperature.

[0023] Step 102: Calculate the stress score for each of the different rendering regions in the digital human based on the behavior complexity, terminal performance score, and preset region priority, and adjust the model accuracy for each of the different rendering regions based on the stress score.

[0024] In the embodiments of this application, the digital human, as a three-dimensional virtual character, can be divided into multiple different rendering areas in the image according to its spatial position and visual importance. For example, the facial features of the digital human can be divided into the core rendering area, the hands, shoulders, and other areas into the key rendering areas, the torso and limbs into the ordinary rendering area, and the background props into the background area.

[0025] Different rendering regions have different requirements for rendering quality. Therefore, it is necessary to pre-set region priorities. Preset region priorities are importance weight values ​​assigned to each rendering region beforehand. With limited rendering resources, preset region priorities determine the priority of each region for rendering. In one example, the priority of the face region can be set to the highest value of 1.0, the hand region to 0.8, the torso region to 0.5, and the background region to 0.2. Preset region priorities can be flexibly configured according to different application scenarios (such as live streaming, customer service, and education), and can also be dynamically adjusted during operation.

[0026] In this application, the stress score is used to characterize the rendering stress experienced by a rendering region under the current time and current terminal performance. The higher the stress score, the less suitable the rendering region is for using a high-precision model, otherwise it will lead to rendering stuttering or frame rate drop; conversely, the lower the stress score, the more computing power the terminal has to support high-precision rendering of the rendering region.

[0027] In this application, model precision refers to the number of vertices or the level of texture detail in a digital human model. High model precision results in higher image quality and a clearer digital human image; medium model precision achieves a balance between image quality and performance; low model precision results in a loss of image detail but faster rendering speed.

[0028] In one example, when adjusting model accuracy, at least one stress score threshold can be set. For instance, when the stress score of a rendering region is higher than a first threshold (e.g., 0.8), it indicates that the current rendering pressure of that region is high, and its model accuracy should be switched to low. When the stress score is between a second threshold and a first threshold (e.g., 0.5 to 0.8), medium model accuracy is used. When the stress score is lower than a second threshold (e.g., 0.5), high-accuracy model is retained.

[0029] Step 103: Adjust the rendering frame rate of different rendering regions according to the behavior complexity and terminal performance score.

[0030] In the embodiments of this application, the rendering frame rate can be measured in frames per second (FPS). A higher rendering frame rate results in a smoother visual experience, but also increases rendering load and network bandwidth consumption. In this application, it is not necessary to use a uniform frame rate for all rendering regions; instead, a separate rendering frame rate is determined for each different rendering region.

[0031] In one example, when adjusting the rendering frame rate for different rendering regions, adjustments can be made based on behavioral complexity and terminal performance score. The higher the behavioral complexity, the more intense the digital human's movements, so the rendering frame rate for key regions needs to be increased; while the lower the terminal performance score, the lower the overall frame rate should be to avoid overload.

[0032] Furthermore, differentiated processing can be implemented based on the importance of each rendering region: a higher rendering frame rate should be maintained for core areas such as the face and hands, while the rendering frame rate can be reduced for background areas. In one possible implementation, a base frame rate (e.g., 30 frames per second) is set for each rendering region, and then the target rendering frame rate for each region is calculated based on the terminal performance score and behavioral complexity, combined with a priority correction factor for each rendering region.

[0033] It should be noted that adjusting the model precision is used to control the amount of computation for each frame of rendering, while adjusting the rendering frame rate is used to control the number of frames rendered per unit of time. Model precision and rendering frame rate can affect the total amount of computation for rendering, but they are two independent dimensions in the rendering process.

[0034] Step 104: Based on the adjusted model accuracy, render different rendering regions according to their respective adjusted rendering frame rates to obtain the rendering results of different rendering regions. Then, merge the rendering results to obtain the rendered video stream.

[0035] In the embodiments of this application, after each rendering region is assigned its own model precision and rendering frame rate, each rendering region is rendered separately during the rendering stage: for a certain rendering region, a local rendering image of that rendering region, i.e., the rendering result, is generated by using the corresponding model precision and rendering frame rate.

[0036] Since the rendering frame rates of different rendering regions may be different, they need to be rendered independently according to their respective time steps, and the completed rendering results are superimposed or blended according to spatial position in each frame to obtain the final rendered image.

[0037] In one feasible implementation, the merging process can be image compositing, applying processing to handle edge transitions and occlusion relationships between different rendering regions. After merging, a rendered video stream is obtained. The rendered video stream can be directly displayed on the screen of the terminal device (edge-side rendering scenario), or it can be pushed to the terminal device for playback via the network (cloud rendering scenario).

[0038] The image rendering method disclosed herein dynamically calculates the stress score of each rendering region to adjust the model accuracy by comprehensively analyzing the behavioral complexity of the real-time actions of the digital human, the performance score of the terminal device, and the preset priority of different rendering regions. Simultaneously, it adjusts the rendering frame rate of each rendering region according to the behavioral complexity and terminal performance, so that the allocation of rendering resources can match the actual needs and terminal capabilities. It can effectively release the rendering load of non-critical regions while ensuring high-quality rendering of critical regions. It solves the problems of high computing power consumption of full-frame and full-region rendering, poor experience in weak network environments, and poor performance of local rendering methods in scheduling complex behaviors. Therefore, this application can improve the smoothness and visual consistency of digital humans, enhance the adaptability to terminals with different performance and network environments, achieve accurate allocation of rendering resources, and improve the presentation quality and user experience of digital humans.

[0039] In one possible implementation of this application embodiment, when obtaining the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, it can be achieved in the following ways, but not limited to: real-time identification of multiple preset action units activated in different rendering regions of the digital human, wherein the preset action units include at least one of basic action units, advanced action units, and complex action units; calculating the weight sum of the multiple preset action units activated in different rendering regions according to the preset weights corresponding to the multiple preset action units, and using the weight sum as the behavior complexity, wherein different rendering regions correspond to different behavior complexities.

[0040] In the embodiments of this application, each rendering area may perform one or more independent actions simultaneously. Therefore, it is necessary to pre-design preset action units that can be automatically recognized by computer vision or skeletal animation systems. A preset action unit refers to the smallest action unit that constitutes the complex behavior of a digital human, and the preset action unit at least includes the movement patterns of various parts of the digital human's body.

[0041] In this application, preset action units can be divided into different categories based on the varying rendering resource consumption of different actions. In one possible implementation, the preset action unit includes at least one of the following: basic action unit, advanced action unit, and complex action unit.

[0042] Basic motion units refer to simple movements with small amplitude, few skeletal points, and low requirements for rendering detail, such as a digitizer blinking (left / right), nodding (up / down), tilting the head (left / right), and a slight upturn of the corners of the mouth. Advanced motion units refer to movements with medium amplitude, involving the coordination of multiple joints, and requiring moderate rendering accuracy, such as waving one hand, shrugging the shoulder, rotating the waist, and simple finger flexion and extension. Complex motion units refer to complex movements with large amplitude, involving the coordination of multiple limbs throughout the body, and requiring high rendering accuracy and frame rate, such as simultaneously performing complex coordinated movements with both hands (e.g., clapping, playing an instrument), rapidly rotating the head beyond a certain angle (e.g., 30 degrees), continuous stepping or jumping, and full-body dance movements.

[0043] By analyzing the skeletal animation data of digital humans or the optical flow information in the video stream in real time, the activated preset motion units can be identified. Activation means that the motion characteristic value of the preset motion unit exceeds the preset threshold. For example, if the head rotation angle is greater than 5 degrees, the head tilt motion unit is activated.

[0044] In this application, each preset action unit is pre-assigned a preset weight. The magnitude of the preset weight indicates the amount of rendering resources consumed by the preset action unit. For example, basic action units have low preset weights because the movements are simple and the requirements for model accuracy are low. For example, the weight of blinking is 0.1 and the weight of nodding is 0.2. Advanced action units involve more limb movements and have moderate rendering requirements, so they have medium preset weights. For example, the weight of waving one hand is 0.6 and the weight of shrugging the shoulder is 0.5. Complex action units involve large-amplitude and multi-limb coordinated movements and consume a lot of rendering resources. Therefore, they have higher preset weights. For example, the weight of coordinated two-handed movements is 0.9, the weight of head rotation exceeding 30 degrees is 0.8, and the weight of continuous stepping is 0.7.

[0045] It should be noted that the specific values ​​of the preset weights can be pre-calibrated and adjusted according to the complexity of different digital human models, application scenarios, and visual importance. This application does not impose specific numerical limitations on these values.

[0046] In one possible embodiment, since different rendering regions may activate different preset action units at the same time, different rendering regions correspond to different behavioral complexities. For example, the digital human's facial region activates preset action units for blinking and head tilting, and the weight sum may be 0.3; while the hand region activates a single-handed wave, and the weight sum is 0.6.

[0047] This application, by assigning different behavioral complexities to different rendering regions, can accurately identify the degree of movement and rendering requirements of various parts of a digital human. Therefore, when calculating the stress score of each rendering region in conjunction with the terminal performance score, more rendering resources can be allocated to rendering regions with high behavioral complexity, or the rendering quality of rendering regions with low behavioral complexity can be reduced first when resources are scarce, thus achieving fine-grained rendering scheduling.

[0048] In one feasible scenario, the calculation of behavioral complexity can also be achieved through, but is not limited to, the following methods: real-time dynamic identification of all preset action units in the current frame of the video stream, and calculation of the behavioral complexity of the current frame based on preset weights. Behavioral complexity The calculation formula is shown in formula (1): Formula (1) in, For behavioral complexity; For the first Whether a preset action unit is activated (0 if not activated, or 1 if activated); For the first Preset weights for each preset action unit; The number of preset action units.

[0049] In one possible implementation of this application embodiment, when obtaining the terminal performance score corresponding to the terminal device, the following methods can be used, but are not limited to: collecting the processor's clock frequency, floating-point performance, and remaining memory capacity in the terminal device; calculating the first ratio of the clock frequency to the first preset maximum value, the second ratio of the floating-point performance to the second preset maximum value, and the third ratio of the remaining memory capacity to the third preset maximum value; multiplying the first ratio, the second ratio, and the third ratio by their respective preset weighting factors and summing them to obtain the terminal performance score.

[0050] In the embodiments of this application, the processor's clock speed refers to the clock speed of the Central Processing Unit (CPU), which represents the number of clock cycles that a CPU core can process per unit of time. Floating-point performance generally refers to the ability of a Graphics Processing Unit (GPU) or a dedicated AI processor to perform floating-point operations. Remaining memory capacity refers to the amount of memory in the terminal device that is not currently occupied by other applications and is available for rendering tasks.

[0051] The first preset maximum value represents the expected value of the terminal device's CPU clock speed; the second preset maximum value represents the expected value of the high-performance GPU's floating-point performance; the third preset maximum value represents the upper limit of sufficient memory capacity when rendering digital humans. Specifically, these values ​​can be adjusted according to the actual situation.

[0052] The calculation of the terminal performance score can also be achieved through, but is not limited to, the following formula (2):

[0053] in, Score the terminal performance; CPU clock speed, For GPU floating-point performance, Remaining memory; The first preset maximum value, The second preset maximum value, The third preset maximum value; The preset weighting factors, such as 0.3, 0.5, and 0.2, can be adjusted according to the actual situation.

[0054] In one possible implementation of this application embodiment, when calculating the stress score corresponding to each of the different rendering regions in the digital human based on behavioral complexity, terminal performance score, and preset region priority, it can be implemented in the following ways, but not limited to: obtaining the action mutation rate corresponding to the digital human, wherein the action mutation rate is used to characterize the degree of action change of the digital human between adjacent frames of the video stream to be rendered; calculating the stress score corresponding to each of the different rendering regions based on the behavioral complexity, preset region priority, terminal performance score, action mutation rate, and preset dynamic adjustment coefficient.

[0055] In the embodiments of this application, in digital human animation or video streams to be rendered, action changes are continuous. For example, a waving action may take multiple frames from start to finish, with minimal differences in action between adjacent frames. However, in some scenarios, the digital human may experience sudden switching or drastic changes in action, such as suddenly starting to wave rapidly from a stationary state, changing from a smile to a loud laugh, or instantly jumping from one pose to another. Sudden changes in action between adjacent frames can impose significant instantaneous stress on the rendering system. Therefore, it is necessary to reflect these sudden changes in action in a timely manner through a stress score; thus, the action mutation rate needs to be considered when calculating the stress score.

[0056] In one possible implementation, the motion abruptness rate can be set to a value between 0 and 1, with a higher value indicating a more drastic change in motion between adjacent frames. Methods for calculating the motion abruptness rate include, but are not limited to: comparing the changes in rotation angles of various skeletal joints of the digital human between the current and previous frames, and comparing the differences in the activated preset motion units between the current and previous frames.

[0057] In one scenario example, if no new preset action units are added between adjacent frames, and the amplitude of any preset action unit changes, the action mutation rate is close to 0; if multiple preset action units with high preset weights are suddenly activated between adjacent frames, the action mutation rate is close to 1.

[0058] In this application, the preset dynamic adjustment coefficient is a configurable weight parameter used to adjust the impact of the action mutation rate on the stress score. The preset dynamic adjustment coefficient can be pre-set according to the specific needs of the application scenario. For example, in a game scenario, the digital human's actions may be more intense, so a higher dynamic adjustment coefficient (such as 0.3) can be set, making the action mutation rate account for a larger proportion of the stress score; while in a film or broadcasting scenario, the digital human's actions are relatively smooth and coherent, so a lower dynamic adjustment coefficient (such as 0.1) can be set, making the stress score mainly determined by the behavioral complexity and terminal performance score.

[0059] In one possible embodiment, when calculating the stress score, the behavior complexity can be multiplied by the logarithmic growth value of the region priority, divided by the sum of the terminal performance score and the unit constant, and then added to the result of the action mutation rate multiplied by a preset dynamic adjustment coefficient to obtain the stress score of the rendering region.

[0060] In one scenario example, the process of calculating a stress score can be implemented in, but is not limited to, the following ways: Define the priority of the preset area You can define predefined region importance weights, such as: face = 1.0, hands = 0.8, background = 0.2; and define the action mutation rate. This represents the degree of drastic change in motion between adjacent frames, with a value ranging from 0 to 1. A larger value indicates a more sudden change.

[0061] The stress score for different rendering regions is calculated in real time based on the behavioral complexity and terminal performance score, as shown in formula (3): Formula (3) Parameter description: For behavioral complexity; Prioritize by region; Rate the terminal performance; The rate of change of action; To dynamically adjust the coefficients, presets can be made based on the scene type, such as: game scenes. =0.3, film and television scene =0.1.

[0062] This application calculates a stress score for each rendering region, which comprehensively reflects the complexity of the current action, the upper limit of the device's capabilities, sudden changes in the action, and the characteristics of the scene. This stress score can be used as a basis for accurately adjusting the model's precision.

[0063] In one possible implementation of this application embodiment, when adjusting the model precision of different rendering regions according to the pressure score, it can be implemented in the following ways, but not limited to: In response to the target pressure score corresponding to the target rendering region being greater than a first threshold, the model precision of the target rendering region is switched to low model precision, wherein the target rendering region is any rendering region in the digital human; In response to the target pressure score corresponding to the target rendering region being no greater than the first threshold but greater than a second threshold, the model precision of the target rendering region is switched to medium model precision; In response to the target pressure score corresponding to the target rendering region being no greater than the second threshold, the model precision of the target rendering region is retained as high model precision; wherein the first threshold is greater than the second threshold, low model precision means that the number of vertices of the model is no greater than a first preset number of vertices, medium model precision means that the number of vertices of the model is greater than the first preset number of vertices but no greater than the second preset number of vertices, and high model precision means that the number of vertices of the model is no less than a third preset number of vertices, and the third preset number of vertices is greater than the second preset number of vertices.

[0064] In the embodiments of this application, the target rendering area can be any one of the following rendering areas: the face area, hand area, torso area, leg area, or background area of ​​the digital human.

[0065] The first threshold is a pre-set pressure judgment threshold used to distinguish between high-pressure and non-high-pressure states. The specific value of this first threshold can be set according to the application scenario, the complexity of the digital human model, and the performance of the terminal device. For example, it can be set to 0.8 (within the pressure score range of 0 to 1). When the target pressure score is greater than the first threshold, it indicates that the target rendering area is under high rendering pressure; therefore, the model precision of the target rendering area needs to be switched to a low model precision.

[0066] The second threshold is another pre-set pressure judgment threshold, which is lower than the first threshold and is used to distinguish between medium and low pressure states. For example, the second threshold can be set to 0.5. When the target pressure score is below the first threshold and above the second threshold (i.e., 0.5 < target pressure score ≤ 0.8), it indicates that the rendering pressure on the target rendering area is at a medium level. At this time, the model accuracy of the target rendering area is switched to medium model accuracy.

[0067] When the target stress score is less than or equal to the second threshold (i.e., ≤0.5), it indicates that the target rendering region is under low rendering stress. Therefore, the model accuracy is preserved or switched to high model accuracy.

[0068] It should be noted that low model precision means that the number of vertices in the model is within a low range, that is, no greater than the first preset number of vertices. The first preset number of vertices is a predefined upper limit value, for example, it can be set to 5000 vertices.

[0069] Medium model precision indicates that the number of vertices in the model falls within the middle range, specifically greater than the first preset number of vertices and not greater than the second preset number of vertices. The second preset number of vertices is a predefined upper limit value, which is greater than the first preset number of vertices; for example, it can be set to 20,000 vertices.

[0070] High model precision means that the number of vertices in the model is in a high range, that is, not less than the third preset number of vertices. The third preset number of vertices is greater than the second preset number of vertices. In other words, the lower limit of the number of vertices in high model precision (the third preset number of vertices) is greater than the upper limit of the number of vertices in medium model precision (the second preset number of vertices). For example, the third preset number of vertices can be set to 100,000 vertices.

[0071] In one scenario example, adjustments to model accuracy can also be achieved in, but are not limited to, the following ways: Based on the stress score, the model precision of different rendering areas is dynamically adjusted. Specifically, the rendering areas mainly include: core area—facial features, fingertips, etc.; key area—shoulders, waist, etc.; normal area—thighs, calves, etc.; background area—scene props, skybox, etc.

[0072] The model accuracy adjustment decision is as follows: High-pressure region (Lr > 0.8): Switch to low-precision model (number of vertices ≤ 5000); Medium pressure region (0.5 < Lr ≤ 0.8): Use medium precision model (number of vertices ≤ 20000); Low-pressure region (Lr≤0.5): Retain high-precision model (number of vertices≥100000).

[0073] This application achieves dynamic and adaptive adjustment of model precision by adjusting the model precision in different regions based on stress scores. Simultaneously, through explicit threshold comparisons (first threshold greater than second threshold) and clear vertex count grading standards (low precision ≤ first preset vertex count < medium precision ≤ second preset vertex count < third preset vertex count ≤ high precision), it can quickly determine the appropriate model precision for each rendering region, thereby reducing the overall rendering load while ensuring image quality in critical areas.

[0074] In one possible implementation of this application embodiment, when adjusting the rendering frame rate of different rendering regions based on behavioral complexity and terminal performance score, the following methods can be used, but are not limited to: obtaining the base frame rate corresponding to each of the different rendering regions; calculating the target rendering frame rate of each of the different rendering regions based on the terminal performance score, preset priority correction coefficient, preset bandwidth correction coefficient, and the behavioral complexity and base frame rate corresponding to each of the different rendering regions; and adjusting the base frame rate corresponding to each of the different rendering regions to the target rendering frame rate.

[0075] In the embodiments of this application, the base frame rate is a pre-set baseline rendering frequency for each rendering region, indicating that without dynamic adjustment, each rendering region is rendered according to the base frame rate by default. Different rendering regions can use the same base frame rate, or different base frame rates can be used according to the importance of the rendering region. For example, the general base frame rate for the entire digital human scene can be set to 30 frames / second; or, the base frame rate for the facial region can be pre-set to 30 frames / second, and the base frame rate for the background region can be set to 15 frames / second. The value of the base frame rate can be preset and configured according to the needs of specific application scenarios, and this application does not impose specific limitations on it.

[0076] The target rendering frame rate refers to the rendering frequency that the rendering area is actually used for rendering after dynamic adjustment.

[0077] The preset priority correction factor is a weighting factor pre-set for each rendering region, used to characterize the visual importance of the corresponding rendering region in the frame rate dimension. Unlike the preset region priority, the preset priority correction factor is used for differentiated frame rate adjustment. For example, a higher preset priority correction factor, such as 1.2, can be assigned to the facial region; a lower preset priority correction factor, such as 0.5, can be assigned to the background region. By using the preset priority correction factor, under the same terminal performance score or behavioral complexity, the frame rate drop in critical regions can be less than that in non-critical regions, thus achieving the allocation of frame rate resources to important regions.

[0078] The preset bandwidth correction factor reflects the impact of current network bandwidth conditions on the rendering frame rate. In cloud rendering mode, the cloud server encodes the rendered footage and pushes it to the terminal device via the network. The network bandwidth affects the video stream bitrate and frame rate. When network bandwidth is sufficient, the preset bandwidth correction factor can be set to a value close to or greater than 1; when the network is in a weak network environment (low or unstable bandwidth), the preset bandwidth correction factor can be set to a value less than 1, thereby actively reducing the rendering frame rate in non-core areas. In terminal device rendering scenarios, this preset bandwidth correction factor can be set to 1, meaning it has no additional impact on the frame rate.

[0079] In one possible embodiment, when determining the target rendering frame rate, the target behavior complexity corresponding to the target rendering region can be added to a first value to obtain a first intermediate value. The ratio of the terminal performance score to the median of the first intermediate value can be calculated. The smaller value between the median ratio and a preset unit constant can be selected as the target value. The target value can be multiplied by the target base frame rate, the target preset priority correction coefficient, and the target preset bandwidth correction coefficient corresponding to the target rendering region to obtain the target rendering frame rate corresponding to the target rendering region. This process continues until the target rendering frame rates of different rendering regions in the digital human are calculated.

[0080] In one scenario example, the process of calculating the target rendering frame rate can be achieved by, but is not limited to, the following formula (4): Formula (4) in, Render the target frame rate. Base frame rate (e.g., 30fps); The terminal performance score is (0-1). The behavioral complexity is (0-10). Set the preset priority adjustment coefficient (e.g., core area weight 1.2, background weight 0.5). This is a preset bandwidth correction factor (to reduce the frame rate in non-core areas when the network is weak).

[0081] This application calculates personalized target rendering frame rates for different rendering regions, enabling the allocation of rendering resources to adapt to the current action complexity, device computing power, and network environment, thereby saving computing and bandwidth resources while ensuring smooth rendering in core areas.

[0082] In one possible implementation of this application embodiment, the different rendering areas include at least a core area, a key area, a normal area, and a background area; The motion mutation rate is determined based on the degree of drastic change in the digital human's motion between adjacent frames; the larger the value, the more drastic the motion change. The preset region priority represents the importance weight of different rendering regions. Among them, the core region has a higher priority than the key region, the key region has a higher priority than the normal region, and the normal region has a higher priority than the background region.

[0083] In the embodiments of this application, the core area refers to the part of the digital human that carries the most facial expression information and interactive semantics, including but not limited to the facial features (such as eyes, mouth, eyebrows, etc.) and the fingertips. The key area refers to areas other than the core area that play an important role in the overall posture and body language expression of the digital human, including but not limited to the shoulders, waist, and upper arms. The ordinary area refers to areas of relatively low visual importance on the digital human's body, including but not limited to the thighs, calves, feet, and parts of the torso. The background area refers to environmental elements in the scene in which the digital human is located, such as: the backdrop of a virtual studio, the skybox, surrounding scene props (tables, chairs, flowers, etc.), and lighting effects.

[0084] In one possible implementation of this application embodiment, when merging the rendering results to obtain a rendered video stream, the following methods can be used, but are not limited to: in response to cloud rendering, merging the rendering results through the cloud to obtain a rendered video stream, and sending the rendered video stream to the terminal device; in response to terminal device rendering, merging the rendering results through the terminal device to obtain a rendered video stream.

[0085] In the embodiments of this application, the model precision of each rendering area is adjusted, and the rendering results of each area (different frame rates) are merged into the final frame. If it is cloud rendering, the cloud pushes the current frame to the terminal device in real time. If it is edge rendering (terminal device rendering), the terminal device directly merges and renders the frames.

[0086] In summary, the embodiments of this application can achieve the following technical effects: 1. By comprehensively analyzing the behavioral complexity of real-time actions of digital humans, the performance score of terminal devices, and the preset priorities of different rendering regions, the stress score of each rendering region is dynamically calculated to adjust the model accuracy. Simultaneously, the rendering frame rate of each rendering region is adjusted according to the behavioral complexity and terminal performance, so that the allocation of rendering resources can match actual needs and terminal capabilities. This can effectively release the rendering load of non-critical regions while ensuring high-quality presentation of key areas. It solves the problems of high computing power consumption of full-frame and full-region rendering, poor experience in weak network environments, and poor performance of local rendering methods for complex behavior scheduling. Therefore, this application can improve the smoothness and visual consistency of digital humans, enhance the adaptability to terminals with different performance and network environments, achieve precise allocation of rendering resources, and improve the presentation quality and user experience of digital humans.

[0087] 2. By modeling the complexity of the current actions and behaviors of the digital human and combining it with the perception of terminal performance levels, regional-level rendering granularity scheduling is carried out, thereby saving rendering resources to the greatest extent while satisfying visual expression.

[0088] 3. Construct a preset weight matrix for digital human motion units, analyze the activation status of basic, advanced and complex actions in the video stream in real time, obtain a behavior complexity score, and use the behavior complexity score as a decision parameter for rendering resource allocation. This breaks through the traditional static hardware-dependent rendering mechanism, realizes control from action semantic understanding to resource scheduling, and improves the rendering adaptability in highly dynamic scenes.

[0089] 4. By integrating behavioral complexity, terminal performance score, regional priority weight, and action mutation rate, a regional stress score is constructed. Based on the stress score, high / medium / low stress regions are dynamically divided, and model accuracy is graded and adjusted to achieve a balance between ensuring image quality in key regions and releasing resources in non-key regions, thereby optimizing rendering efficiency under complex actions.

[0090] 5. By adjusting the rendering frame rate based on behavioral complexity and terminal performance, the system prioritizes frame rate protection in the core area and reduces frame rate in the non-core area based on bandwidth sensitivity. This resolves the conflict between image quality, smoothness, and bandwidth resources in cloud rendering scenarios and ensures the continuity of core interactive experience in weak network environments.

[0091] Corresponding to the image rendering method described above, the present invention also proposes an image rendering apparatus. Since the apparatus embodiments of the present invention correspond to the method embodiments described above, details not disclosed in the apparatus embodiments can be referred to in the method embodiments described above, and will not be repeated here.

[0092] Figure 2 This is a schematic diagram of the structure of an image rendering apparatus provided in an embodiment of the present disclosure, such as... Figure 2 As shown, it includes: The acquisition unit 21 is used to acquire the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, and to acquire the terminal performance score corresponding to the terminal device, which is used to render the digital human. The calculation unit 22 is used to calculate the stress score corresponding to each of the different rendering areas in the digital human based on the behavior complexity, terminal performance score and preset area priority. The first adjustment unit 23 is used to adjust the model accuracy of different rendering areas according to the pressure score; The second adjustment unit 24 is used to adjust the rendering frame rate of different rendering regions according to the behavior complexity and terminal performance score. Rendering unit 25 is used to render different rendering regions according to their respective adjusted rendering frame rates based on the adjusted model accuracy, so as to obtain the rendering results of different rendering regions. Merging unit 26 is used to merge the rendering results to obtain the rendered video stream.

[0093] Furthermore, in one possible implementation of this disclosure embodiment, the acquisition unit 21 is specifically used for: Real-time identification of multiple preset action units activated in different rendering areas of the digital human. The preset action units include at least one of basic action units, advanced action units, and complex action units. Based on the preset weights corresponding to each of the multiple preset action units, the weights of the multiple preset action units activated in different rendering regions are calculated, and the weight sum is used as the behavior complexity. Different rendering regions correspond to different behavior complexities.

[0094] Furthermore, in one possible implementation of this disclosure embodiment, the acquisition unit 21 is specifically used for: Collect the processor's clock speed, floating-point performance, and remaining memory capacity in the terminal device; Calculate the first ratio of the main frequency to the first preset maximum value, the second ratio of the floating-point performance to the second preset maximum value, and the third ratio of the remaining memory capacity to the third preset maximum value, respectively. The terminal performance score is obtained by multiplying the first ratio, the second ratio, and the third ratio by their respective preset weighting factors and summing them.

[0095] Furthermore, in one possible implementation of this disclosure embodiment, the computing unit 22 is specifically used for: Obtain the motion mutation rate corresponding to the digital human, whereby the motion mutation rate is used to characterize the degree of motion change of the digital human between adjacent frames of the video stream to be rendered; The stress score for each rendering region is calculated based on its corresponding behavioral complexity, preset region priority, terminal performance score, action mutation rate, and preset dynamic adjustment coefficient.

[0096] Furthermore, in one possible implementation of this disclosure, the first adjustment unit 23 is specifically used for: In response to the target pressure score corresponding to the target rendering region being greater than the first threshold, the model precision of the target rendering region is switched to low model precision, where the target rendering region is any rendering region in the digital human; If the target pressure score corresponding to the target rendering region is not greater than the first threshold but is greater than the second threshold, the model precision of the target rendering region is switched to medium model precision. If the target pressure score corresponding to the target rendering region is not greater than the second threshold, the model accuracy of the target rendering region is retained as high model accuracy. Among them, the first threshold is greater than the second threshold, low model precision means that the number of vertices in the model is not greater than the first preset number of vertices, medium model precision means that the number of vertices in the model is greater than the first preset number of vertices and not greater than the second preset number of vertices, and high model precision means that the number of vertices in the model is not less than the third preset number of vertices, and the third preset number of vertices is greater than the second preset number of vertices.

[0097] Furthermore, in one possible implementation of this embodiment, the second adjustment unit 24 is specifically used for: Obtain the base frame rate corresponding to each of the different rendering regions; Based on the terminal performance score, preset priority correction coefficient, preset bandwidth correction coefficient, and the behavioral complexity and base frame rate of each rendering region, calculate the target rendering frame rate of each rendering region. Adjust the base frame rate corresponding to each of the different rendering regions to the target rendering frame rate.

[0098] Furthermore, in one possible implementation of this disclosure embodiment, the different rendering areas include at least a core area, a key area, a normal area, and a background area; The motion mutation rate is determined based on the degree of drastic change in the digital human's motion between adjacent frames; the larger the value, the more drastic the motion change. The preset region priority represents the importance weight of different rendering regions. Among them, the core region has a higher priority than the key region, the key region has a higher priority than the normal region, and the normal region has a higher priority than the background region.

[0099] Furthermore, in one possible implementation of this disclosure, the merging unit 26 is specifically used for: In response to the rendering method being cloud rendering, the rendering results are merged and processed in the cloud to obtain the rendered video stream, which is then sent to the terminal device. In response to the rendering method being terminal device rendering, the rendering results are merged and processed by the terminal device to obtain the rendered video stream.

[0100] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0101] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0102] Figure 3A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0103] like Figure 3 As shown, device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 302 or a computer program loaded from storage unit 308 into RAM (Random Access Memory) 303. RAM 303 may also store various programs and data required for the operation of device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. I / O (Input / Output) interface 305 is also connected to bus 304.

[0104] Multiple components in device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of monitors, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0105] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as image rendering methods. For example, in some embodiments, the image rendering method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform the aforementioned image rendering method by any other suitable means (e.g., by means of firmware).

[0106] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0107] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0108] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0109] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0110] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0111] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0112] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0113] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0114] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An image rendering method, characterized in that, include: The behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered is obtained, and the terminal performance score corresponding to the terminal device is obtained, wherein the terminal device is used to render the digital human. Based on the behavioral complexity, the terminal performance score, and the preset region priority, calculate the stress score corresponding to each of the different rendering regions in the digital human, and adjust the model accuracy of each of the different rendering regions according to the stress score. The rendering frame rate of each of the different rendering regions is adjusted based on the behavioral complexity and the terminal performance score. Based on the adjusted model accuracy, the different rendering regions are rendered according to their respective adjusted rendering frame rates to obtain the rendering results of the different rendering regions. The rendering results are then merged to obtain the rendered video stream.

2. The method according to claim 1, characterized in that, The behavioral complexity of obtaining the real-time action behavior of the digital human in the video stream to be rendered includes: The system can identify in real time multiple preset action units activated in different rendering areas of the digital human. The preset action units include at least one of basic action units, advanced action units, and complex action units. Based on the preset weights corresponding to the multiple preset action units, the weight sum of the multiple preset action units activated in each of the different rendering regions is calculated, and the weight sum is used as the behavior complexity, wherein different rendering regions correspond to different behavior complexities.

3. The method according to claim 1, characterized in that, The method of obtaining the terminal performance score corresponding to the terminal device includes: Collect the processor's clock speed, floating-point performance, and remaining memory capacity in the terminal device; Calculate the first ratio of the main frequency to the first preset maximum value, the second ratio of the floating-point performance to the second preset maximum value, and the third ratio of the remaining memory capacity to the third preset maximum value, respectively. The terminal performance score is obtained by multiplying the first ratio, the second ratio, and the third ratio by their respective preset weighting factors and summing them.

4. The method according to claim 1, characterized in that, The step of calculating the stress score corresponding to different rendering regions in the digital human based on the behavior complexity, the terminal performance score, and the preset region priority includes: Obtain the motion mutation rate corresponding to the digital human, wherein the motion mutation rate is used to characterize the degree of motion change of the digital human between adjacent frames of the video stream to be rendered; The stress score for each of the different rendering regions is calculated based on the behavioral complexity of each region, the preset region priority, the terminal performance score, the action mutation rate, and the preset dynamic adjustment coefficient.

5. The method according to claim 1, characterized in that, The step of adjusting the model accuracy of different rendering regions based on the pressure score includes: In response to the target pressure score corresponding to the target rendering region being greater than a first threshold, the model precision of the target rendering region is switched to a low model precision, wherein the target rendering region is any rendering region in the digital human; In response to the target pressure score corresponding to the target rendering region being no greater than the first threshold and greater than the second threshold, the model precision of the target rendering region is switched to medium model precision. In response to the target pressure score corresponding to the target rendering region not being greater than the second threshold, the model accuracy of the target rendering region is retained as high model accuracy; Wherein, the first threshold is greater than the second threshold, the low model precision means that the number of vertices in the model is not greater than the first preset number of vertices, the medium model precision means that the number of vertices in the model is greater than the first preset number of vertices and not greater than the second preset number of vertices, and the high model precision means that the number of vertices in the model is not less than the third preset number of vertices, and the third preset number of vertices is greater than the second preset number of vertices.

6. The method according to claim 1, characterized in that, The step of adjusting the rendering frame rate of each of the different rendering regions based on the behavior complexity and the terminal performance score includes: Obtain the base frame rate corresponding to each of the different rendering regions; Based on the terminal performance score, the preset priority correction coefficient, the preset bandwidth correction coefficient, and the behavioral complexity and base frame rate corresponding to each of the different rendering regions, calculate the target rendering frame rate for each of the different rendering regions. The base frame rate corresponding to each of the different rendering regions is adjusted to the target rendering frame rate.

7. The method according to claim 4, characterized in that, The different rendering areas include at least a core area, a key area, a normal area, and a background area; The motion mutation rate is determined based on the degree of drastic change in the digital human's motion between adjacent frames; a larger value indicates a more drastic change in motion. The preset region priority represents the importance weight of each of the different rendering regions, wherein the core region has a higher priority than the key region, the key region has a higher priority than the normal region, and the normal region has a higher priority than the background region.

8. The method according to claim 1, characterized in that, The step of merging the rendering results to obtain the rendered video stream includes: In response to the rendering method being cloud rendering, the rendering results are merged and processed in the cloud to obtain the rendered video stream, and the rendered video stream is sent to the terminal device; In response to the rendering method being terminal device rendering, the rendering results are merged and processed by the terminal device to obtain the rendered video stream.

9. An image rendering apparatus, characterized in that, include: The acquisition unit is used to acquire the behavior complexity corresponding to the real-time action behavior of the digital human in the video stream to be rendered, and to acquire the terminal performance score corresponding to the terminal device, wherein the terminal device is used to render the digital human. The calculation unit is used to calculate the stress score corresponding to each of the different rendering regions in the digital human based on the behavior complexity, the terminal performance score and the preset region priority. The first adjustment unit is used to adjust the model accuracy of the different rendering regions according to the pressure score; The second adjustment unit is used to adjust the rendering frame rate of each of the different rendering regions according to the behavior complexity and the terminal performance score. The rendering unit is used to render the different rendering regions according to their respective adjusted rendering frame rates based on the adjusted model accuracy, so as to obtain the rendering results of the different rendering regions. The merging unit is used to merge the rendering results to obtain the rendered video stream.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.