CN117523060B - Image quality processing method, device, equipment and storage medium for metauniverse digital person - Google Patents

Image quality processing method, device, equipment and storage medium for metauniverse digital person Download PDF

Info

Publication number
CN117523060B
CN117523060B CN202410008023.0A CN202410008023A CN117523060B CN 117523060 B CN117523060 B CN 117523060B CN 202410008023 A CN202410008023 A CN 202410008023A CN 117523060 B CN117523060 B CN 117523060B
Authority
CN
China
Prior art keywords
image quality
parameter
quality processing
parameters
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202410008023.0A
Other languages
Chinese (zh)
Other versions
CN117523060A (en
Inventor
车守刚
吴湛
段会亮
刘永逵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Virtual Reality Shenzhen Intelligent Technology Co ltd
Original Assignee
Virtual Reality Shenzhen Intelligent Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Virtual Reality Shenzhen Intelligent Technology Co ltd filed Critical Virtual Reality Shenzhen Intelligent Technology Co ltd
Priority to CN202410008023.0A priority Critical patent/CN117523060B/en
Publication of CN117523060A publication Critical patent/CN117523060A/en
Application granted granted Critical
Publication of CN117523060B publication Critical patent/CN117523060B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T15/00—3D [Three Dimensional] image rendering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T19/00—Manipulating 3D models or images for computer graphics
    • G06T19/20—Editing of 3D images, e.g. changing shapes or colours, aligning objects or positioning parts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Graphics (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Databases & Information Systems (AREA)
  • Architecture (AREA)
  • Computer Hardware Design (AREA)
  • Image Processing (AREA)

Abstract

The application relates to the technical field of metauniverse, and discloses a method, a device, equipment and a storage medium for processing image quality of metauniverse digital people. The method comprises the following steps: carrying out population initialization on image quality processing parameters of the target element universe digital person to obtain a plurality of first image quality processing parameters; calculating target fitness data and dynamically adjusting cosine adjustment factors; constructing nonlinear oscillation adjustment factors and performing parameter searching to generate a plurality of second image quality processing parameters; performing image quality rendering simulation operation to obtain target image quality rendering parameters, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features; inputting the target multi-mode fusion characteristics into a cascading forest model for image quality rendering characteristic analysis to obtain image quality rendering characteristic data; and carrying out optimization analysis on the plurality of second image quality processing parameters to obtain target image quality processing parameters, thereby improving the image quality processing accuracy of the metauniverse digital person.

Description

Image quality processing method, device, equipment and storage medium for metauniverse digital person
Technical Field
The application relates to the technical field of metauniverse, in particular to a method, a device, equipment and a storage medium for processing image quality of metauniverse digital people.
Background
With the rapid development of the metauniverse, the fidelity of digital people puts higher demands on user experience and interactivity. Image quality processing of digital people is an important direction in metaspace technology research. The conventional rendering method is difficult to meet the requirements of realistic sensation and multi-modal expression of metauniverse digital people, and therefore, an innovative image quality processing method is needed to improve the visual quality and perceived reality of digital people. The current image quality processing method of digital people mostly depends on static parameter adjustment, and comprehensive consideration of dynamic scenes and multi-mode situations is lacking.
However, the existing research still has some problems in the field of image quality processing of metauniverse digital people. The conventional rendering technology cannot effectively process multi-modal expression of a digital person in a complex scene, so that the generated digital person lacks realism. Secondly, the image quality processing of the current method in a dynamic scene still has a great challenge, because the traditional parameter adjustment cannot flexibly adapt to the changing environment.
Disclosure of Invention
The application provides a method, a device, equipment and a storage medium for processing the image quality of a metauniverse digital person, thereby improving the accuracy of processing the image quality of the metauniverse digital person.
The first aspect of the present application provides a method for processing the image quality of a metauniverse digital person, the method for processing the image quality of the metauniverse digital person comprising:
Carrying out population initialization on image quality processing parameters of a target element universe digital person by adopting a sine and cosine algorithm to obtain an initialization parameter population, wherein the initialization parameter population comprises: a plurality of first image quality processing parameters;
Respectively calculating target fitness data of each first image quality processing parameter, and dynamically adjusting a cosine adjustment factor according to the target fitness data;
Constructing a nonlinear oscillation adjustment factor of the initialization parameter population according to the cosine adjustment factor, and carrying out parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factor to generate a plurality of second image quality processing parameters;
Performing image quality rendering simulation operation on the target meta-universe digital person according to the second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter;
Inputting the target multi-mode fusion characteristics into a preset cascade forest model for image quality rendering characteristic analysis to obtain image quality rendering characteristic data of each second image quality processing parameter;
And carrying out optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering characteristic data to obtain target image quality processing parameters.
A second aspect of the present application provides an image quality processing apparatus for a metauniverse digital person, the image quality processing apparatus comprising:
the initialization module is used for carrying out population initialization on image quality processing parameters of the target universe digital person by adopting a sine and cosine algorithm to obtain an initialization parameter population, wherein the initialization parameter population comprises: a plurality of first image quality processing parameters;
The computing module is used for respectively computing target fitness data of each first image quality processing parameter and dynamically adjusting cosine adjustment factors according to the target fitness data;
The construction module is used for constructing nonlinear oscillation adjustment factors of the initialization parameter population according to the cosine adjustment factors, and carrying out parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factors to generate a plurality of second image quality processing parameters;
The extraction module is used for performing image quality rendering simulation operation on the target meta-universe digital person according to the plurality of second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter;
the analysis module is used for inputting the target multi-mode fusion characteristics into a preset cascade forest model to perform image quality rendering characteristic analysis to obtain image quality rendering characteristic data of each second image quality processing parameter;
and the optimization module is used for carrying out optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering characteristic data to obtain target image quality processing parameters.
A third aspect of the present application provides a computer apparatus comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the computer device to perform the above-described method of image quality processing of a metauniverse digital person.
A fourth aspect of the present application provides a computer-readable storage medium having instructions stored therein, which when run on a computer, cause the computer to perform the above-described method of image quality processing of a metauniverse digital person.
According to the technical scheme provided by the application, the multilevel optimization of the initialization parameter population is realized through the dynamic adjustment of the sine and cosine algorithm and the cosine adjustment factor, and the breadth and the depth of the search space are improved. And the nonlinear oscillation adjustment factor is utilized for parameter searching, so that the flexibility of an algorithm is effectively increased, and the characteristic of image quality processing parameters can be more comprehensively captured. The multi-mode fusion feature extraction is carried out by adopting a convolution long-short-time memory network, a threshold circulation network and a linear regression layer, so that the multi-mode fusion feature extraction is beneficial to comprehensively considering the image quality information in many aspects such as texture, depth scene, color saturation and the like. The feature extraction mode enables the system to be more suitable for the complex image quality requirements of metauniverse digital people, and improves the global perception capability of image quality rendering. And a cascading forest model is introduced to conduct image quality rendering feature analysis, and decision analysis is conducted through a plurality of decision trees, so that different image quality processing situations can be understood and processed more carefully. The use of the model improves the decision accuracy and the robustness of the system, so that the image quality processing is more intelligent and adaptive. And a mechanism for dynamically adjusting the cosine adjustment factor is adopted, and real-time adjustment is carried out according to the target fitness data, so that the adaptability of the algorithm to different image quality scenes is improved. The adaptivity enables the system to be more robust when processing different digital human image quality, and can cope with diversified metauniverse scenes and demands. The searching and selecting efficiency of the target image quality processing parameters is improved through comprehensive optimization strategies such as division, propagation, crossing and variation of a plurality of parameter populations. The comprehensive optimization strategy is beneficial to finding out better image quality processing parameters in a large-scale parameter space, so that the overall image quality rendering effect is improved, and the image quality processing accuracy of the metauniverse digital person is further improved.
Drawings
FIG. 1 is a diagram showing an example of a method for processing image quality of a metauniverse digital person according to an embodiment of the present application;
fig. 2 is a schematic diagram of an image quality processing apparatus for a metauniverse digital person according to an embodiment of the present application.
Detailed Description
The embodiment of the application provides a method, a device, equipment and a storage medium for processing the image quality of a metauniverse digital person, thereby improving the accuracy of processing the image quality of the metauniverse digital person.
The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
For easy understanding, the following describes a specific flow of an embodiment of the present application, referring to fig. 1, and an embodiment of a method for processing image quality of a metauniverse digital person in an embodiment of the present application includes:
Step 101, initializing a population of image quality processing parameters of a target element universe digital person by adopting a sine and cosine algorithm to obtain an initialized parameter population, wherein the initialized parameter population comprises: a plurality of first image quality processing parameters;
it should be understood that the execution subject of the present application may be an image quality processing device of a metauniverse digital person, or may be a terminal or a server, and is not limited in particular herein. The embodiment of the application is described by taking a server as an execution main body as an example.
Specifically, an image quality processing parameter space including a plurality of image quality processing parameters and corresponding parameter intervals is defined. This parameter space covers the important factors of texture map parameters, depth scene parameters, and color saturation parameters, which together determine the final visual presentation effect of the metauniverse of digital people. For example, texture map parameters focus on the richness and realism of details, depth scene parameters affect the spatial and hierarchical sense of objects in the scene, while color saturation parameters directly affect the vividness and contrast of vision. The image quality processing parameter space is operated by a sine and cosine algorithm, and a random value which is uniformly distributed is generated for each image quality processing parameter. The randomness and the global searching capability of the sine and cosine algorithm are utilized to ensure that each parameter can obtain a representative initial value in a defined parameter interval. This approach not only increases the diversity of the parameter search, but also provides a comprehensive and balanced starting point for the next optimization process. And forming an initialization parameter population of the target metauniverse digital person according to the randomly generated image quality processing parameter values. The population includes a plurality of first image quality processing parameters, each of which is a configuration of different aspects of texture mapping, depth scene, and color saturation.
102, Respectively calculating target fitness data of each first image quality processing parameter, and dynamically adjusting a cosine adjustment factor according to the target fitness data;
Specifically, a range of cosine adjustment factors is determined, which is set between 0 and 1, where 0 indicates no adjustment and 1 indicates the greatest degree of adjustment. This range is set to ensure flexibility and fineness of adjustment, allowing the algorithm to flexibly adjust the influence of the parameters according to specific fitness conditions. The calculation of the target fitness data is performed for each of the first image quality processing parameters. And evaluating the image quality of the digital character image under each parameter setting, wherein the image quality comprises a plurality of aspects such as definition of textures, accuracy of colors, depth feeling of a scene and the like. Meanwhile, preset standard fitness data are acquired, and the data serve as reference standards to help judge the advantages and disadvantages of the current parameter setting. Difference data between the target fitness data and the standard fitness data is calculated. Through such a difference calculation, the performance of each first image quality processing parameter with respect to the preset standard can be clearly understood, and such comparison can intuitively reflect which parameters need to be adjusted, and the direction and degree of adjustment. And calculating average difference data corresponding to the target difference data of each first image quality processing parameter, and dynamically adjusting the cosine adjustment factor of each parameter according to the average difference data and the determined value range. The adjustment is based on the size of the difference data, and if the target difference data of a certain parameter is larger, which indicates that the current setting of the certain parameter deviates from the ideal state greatly, the corresponding cosine adjustment factor is adjusted to be closer to 1 so as to adjust the parameter more strongly in the subsequent optimization process. Conversely, if the difference data is small, indicating that the parameter setting has been relatively close to ideal, then the remaining string adjustment factors will be adjusted closer to 0 to maintain the current parameter setting.
Step 103, constructing a nonlinear oscillation adjustment factor of the initialization parameter population according to the cosine adjustment factor, and carrying out parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factor to generate a plurality of second image quality processing parameters;
It should be noted that, according to a preset Logistic chaotic mapping function, the cosine adjustment factors are subjected to chaotic mapping to generate nonlinear oscillation adjustment factors of each first image quality processing parameter. Logistic chaotic mapping is a dynamic system that can produce complex and unpredictable behavior in an iterative process, which makes it well suited for increasing the randomness and heuristics of the search process, thereby avoiding premature trapping of the algorithm into local optima. After the processing of the Logistic chaotic mapping function, each first image quality processing parameter can obtain a nonlinear oscillation adjustment factor corresponding to the first image quality processing parameter. These adjustment factors not only reflect the current state of each parameter, but also imply the direction and magnitude that need to be explored in the search process. And determining the parameter search space of the initialized parameter population and the iteration times thereof according to the nonlinear oscillation adjustment factors. The determination of the parameter search space involves evaluating the variation range and direction of each parameter, and the setting of the iteration times is based on the complexity of the optimization process and the requirement of the solving precision, so that the search process is ensured not to be too lengthy, and the final optimization effect is not affected due to insufficient iteration times. And carrying out parameter searching on the initialized parameter population according to the determined parameter searching space and iteration times. Each iteration adjusts the value of the parameter according to the nonlinear oscillation adjustment factor, such adjustment is dynamic and adaptive, and the search strategy can be flexibly adjusted according to the current search state. And gradually approaching the optimal solution along with the iteration, and finally forming a plurality of second image quality processing parameters. These second image quality processing parameters represent the results of the effective adjustment and optimization of the original parameters under the current optimization framework, which will be used for the subsequent image quality rendering simulation and evaluation to ensure that the achieved image quality effect can meet the high standard requirements of the metauniverse digital man in visual performance.
104, Respectively performing image quality rendering simulation operation on the target meta-universe digital person according to the plurality of second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter;
Specifically, each second image quality processing parameter is input into a preset image quality processing parameter prediction model. The model is a complex neural network including a convolutional long short-term memory network (ConvLSTM), a threshold-cycling network (GRU), and a linear regression layer. The ConvLSTM network performs parameter feature space mapping on each image quality processing parameter so that the parameter can be converted into corresponding parameter feature vectors, and the vectors can effectively characterize the characteristics of original parameters and potential influences of the original parameters on the image quality. These parametric feature vectors are input to the GRU network for feature encoding. The GRU network further extracts and codes information contained in the parameter feature vector to generate a parameter code vector. This translates the original parametric features into a form more suitable for image quality rendering simulation. And inputting the parameter coding vectors into a linear regression layer for image quality rendering simulation operation, so as to obtain target image quality rendering parameters of each second image quality processing parameter. The linear regression layer converts the encoded vectors into specific image quality rendering parameters. In order to further improve accuracy and fineness of image quality rendering, multi-modal parameter classification is performed on target image quality rendering parameters, and a plurality of digital human modal data are generated. By decomposing the image quality rendering parameters into different modes, such as texture, color, light and shadow, the influence of the different parameters on the image quality can be more comprehensively captured. After a plurality of digital human modal data are generated, key features of the modal data are extracted through a feature extraction technology to form a plurality of digital human modal features. These features reflect the effects of the image quality processing parameters in different modalities and provide the necessary information for final image quality assessment. The digital person features of different modes are combined through a feature fusion technology, so that the target multi-mode fusion feature of each second image quality processing parameter is obtained, the image quality processing of the metauniverse digital person is ensured to consider not only the comprehensive effect of each parameter, but also the interaction of different visual modes.
Step 105, inputting the target multi-mode fusion characteristics into a preset cascade forest model for image quality rendering characteristic analysis to obtain image quality rendering characteristic data of each second image quality processing parameter;
Specifically, the target multi-mode fusion characteristic is input into a preset cascade forest model, the model mainly comprises a plurality of decision trees, each decision tree is an independent analysis unit, and the input multi-mode fusion characteristic can be processed and analyzed from different angles. And carrying out careful image quality processing decision analysis on the target multi-modal fusion characteristics through a plurality of decision trees. Each decision tree interprets and analyzes the input characteristic data through the unique decision logic and parameter setting, and generates an initial decision evaluation index. These metrics reflect the understanding and processing effect of each decision tree on the image quality rendering characteristics. Weight data corresponding to each decision tree is obtained. These weight data are set based on the importance and reliability of each decision tree in the image quality processing decision analysis, which helps balance and integrate the output results of the individual decision trees. And carrying out evaluation index weighted analysis on the initial decision evaluation index of each decision tree according to the obtained weight data. The evaluation index of each decision tree is adjusted and recombined according to the corresponding weight, so that more accurate and comprehensive image quality rendering characteristic data is generated. Through the multi-decision tree structure and the weighted analysis mechanism of the cascade forest model, analysis results of different angles can be effectively synthesized, and the utilization efficiency and accuracy of the multi-mode fusion features are improved. The finally obtained image quality rendering characteristic data of each second image quality processing parameter not only synthesizes the analysis results of a plurality of visual angles, but also realizes the optimization and refining of the results through a weight weighting process.
And 106, performing optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering characteristic data to obtain target image quality processing parameters.
Specifically, the acquired image quality rendering feature data is compared with two set evaluation index thresholds, namely a first evaluation index threshold and a second evaluation index threshold. The two thresholds are set to distinguish between different levels of image quality rendering effects, wherein the first evaluation index threshold is lower than the second evaluation index threshold, thereby forming three different parameter classifications. Specifically, when certain image quality rendering characteristic data is lower than the first evaluation index threshold, it is indicated that the corresponding second image quality processing parameters are poorly represented, and therefore, these parameters are classified into the first parameter population. If the image quality rendering characteristic data is between the first evaluation index threshold and the second evaluation index threshold, it is indicated that the performance of these parameters is at a medium level and is therefore classified as a second parameter population. If the image quality rendering characteristic data exceeds a second evaluation index threshold, it is indicated that the corresponding second image quality processing parameters perform well, and these parameters are then divided into a third parameter population. Different optimization strategies are adopted for the three different parameter populations. For the first and second parameter populations, breeding, crossover and mutation operations are performed. The breeding operation aims at increasing the diversity of parameter populations, the crossover operation aims at combining the advantages of different parameters, and the mutation operation is used for introducing new properties so as to explore more unknown optimization directions. For the third parameter population, only the crossover and mutation operations are performed, aiming at further optimizing the parameters which have been well represented, since they have been shown to have a better image quality rendering effect. Through the above steps, a plurality of candidate image quality processing parameters are generated. The parameters are screened and optimized layer by layer, so that the image quality processing effect is achieved or exceeded by higher performance. And (3) carrying out optimization analysis on the candidate image quality processing parameters, and selecting a parameter combination which can improve the image quality of the metauniverse digital person, namely the target image quality processing parameters. This ensures that the finally selected parameters are not only theoretically efficient, but also that in practical applications an optimal image quality rendering effect can be produced. Through the optimization analysis, the image quality of the metauniverse digital person can be ensured to be optimized in the aspects of detail processing, color presentation, texture expression and the like, so that high-fidelity and attractive visual experience is provided in the metauniverse environment.
In the embodiment of the application, the multilevel optimization of the initialization parameter population is realized through the dynamic adjustment of the sine and cosine algorithm and the cosine adjustment factor, and the breadth and the depth of the search space are improved. And the nonlinear oscillation adjustment factor is utilized for parameter searching, so that the flexibility of an algorithm is effectively increased, and the characteristic of image quality processing parameters can be more comprehensively captured. The multi-mode fusion feature extraction is carried out by adopting a convolution long-short-time memory network, a threshold circulation network and a linear regression layer, so that the multi-mode fusion feature extraction is beneficial to comprehensively considering the image quality information in many aspects such as texture, depth scene, color saturation and the like. The feature extraction mode enables the system to be more suitable for the complex image quality requirements of metauniverse digital people, and improves the global perception capability of image quality rendering. And a cascading forest model is introduced to conduct image quality rendering feature analysis, and decision analysis is conducted through a plurality of decision trees, so that different image quality processing situations can be understood and processed more carefully. The use of the model improves the decision accuracy and the robustness of the system, so that the image quality processing is more intelligent and adaptive. And a mechanism for dynamically adjusting the cosine adjustment factor is adopted, and real-time adjustment is carried out according to the target fitness data, so that the adaptability of the algorithm to different image quality scenes is improved. The adaptivity enables the system to be more robust when processing different digital human image quality, and can cope with diversified metauniverse scenes and demands. The searching and selecting efficiency of the target image quality processing parameters is improved through comprehensive optimization strategies such as division, propagation, crossing and variation of a plurality of parameter populations. The comprehensive optimization strategy is beneficial to finding out better image quality processing parameters in a large-scale parameter space, so that the overall image quality rendering effect is improved, and the image quality processing accuracy of the metauniverse digital person is further improved.
In a specific embodiment, the process of executing step 101 may specifically include the following steps:
(1) Defining an image quality processing parameter space, wherein the image quality processing parameter space comprises a plurality of image quality processing parameters and parameter intervals corresponding to each image quality processing parameter, and the plurality of image quality processing parameters comprise: texture map parameters, depth scene parameters, and color saturation parameters;
(2) Respectively generating random values of uniform distribution of each image quality processing parameter in an image quality processing parameter space through a sine and cosine algorithm;
(3) Generating an initialization parameter population of the target metauniverse digital person according to the random value of each image quality processing parameter, wherein the initialization parameter population comprises the following components: a plurality of first image quality processing parameters.
Specifically, an image quality processing parameter space including a plurality of image quality processing parameters and corresponding parameter intervals is defined. In this space, a number of key factors are included, such as texture map parameters, depth scene parameters, and color saturation parameters, which together determine the visual effect of a metauniverse of digital people. For example, texture map parameters mainly affect the surface detail and realism of a digital character image, depth scene parameters focus on the spatial depth and layering of the scene, and color saturation parameters directly affect the color richness and visual impact of the image. A sine and cosine algorithm is used to generate a uniformly distributed random value for each parameter in the image quality processing parameter space. The sine and cosine algorithm is an efficient optimization method that can produce random but uniformly distributed values over a range of values of the parameters, thereby providing a diverse and comprehensive starting point for the next optimization. In this way, it is ensured that each image quality processing parameter has the opportunity to be explored and optimized, without being limited to a narrow range. Then, an initialization parameter population of the target metauniverse digital person is generated according to the random values of each image quality processing parameter. The population includes a plurality of first image quality processing parameters, each of which is an implementation of the image quality processing parameters. For example, for texture map parameters, there may be parameter values for different texture densities and sharpness levels; for depth scene parameters, parameter values of different visual angle depths and light shadow effects are included; for color saturation parameters, then, parameter values of different levels of color saturation from low to high are covered. And generating an initialization parameter population of the target metauniverse digital person according to the random value of each image quality processing parameter, namely a parameter set comprising a plurality of different texture maps, depth scenes and color saturation combinations.
In a specific embodiment, the process of executing step 102 may specifically include the following steps:
(1) Determining a value range of a cosine adjusting factor, wherein 0 represents no adjustment, and 1 represents the maximum adjustment;
(2) Respectively calculating target fitness data of each first image quality processing parameter, and acquiring preset standard fitness data;
(3) Calculating difference data between the target fitness data and the standard fitness data to obtain target difference data of each first image quality processing parameter;
(4) And calculating average difference data corresponding to the target difference data of each first image quality processing parameter, and dynamically adjusting cosine adjustment factors of each first image quality processing parameter according to the average difference data and the value range.
Specifically, the range of the cosine adjustment factor is determined, and this range is set between 0 and 1. Within this range, 0 indicates that the corresponding image quality processing parameter is not subjected to any adjustment, and 1 indicates that the parameter is subjected to the greatest degree of adjustment. This setting allows for fine and flexible adjustment of the image quality processing parameters, ensuring the accuracy and effectiveness of the optimization process. Target fitness data is calculated for each of the first image quality processing parameters. The target fitness data is determined by evaluating the contribution of each parameter to the image quality improvement effect. This evaluation takes into account how the parameters affect the sharpness, color saturation, texture details, etc. of the final image. For example, for texture map parameters, target fitness data is calculated based on its degree of improvement in image detail and realism; for the color saturation parameter, it is evaluated based on its enhancement to the image color richness and visual impact. And then, acquiring preset standard fitness data. These data represent the effect that each parameter should achieve under ideal conditions and can be used as a benchmark for evaluating the fitness data of the target. For example, the standard adaptation data indicates how the best texture map parameters should promote the realism of the image, or how the best color saturation parameters should enhance the visual appeal of the image. And obtaining target difference data of each first image quality processing parameter by calculating the difference between the target fitness data and the standard fitness data. This difference data reveals the deviation of the current parameter setting from the ideal state, providing a basis for subsequent adjustment. For example, if the target fitness data for a particular texture map parameter is far below the standard fitness data, this indicates that the parameter is not sufficiently improving the realism of the image at the current setting and that an adjustment is required. And dynamically adjusting cosine adjustment factors of each parameter according to the average difference data and the value range by calculating the average difference data corresponding to the target difference data of each first image quality processing parameter. This adjustment process is dynamic and the degree of adjustment is determined based on the size of the difference data. For those parameters for which the target difference data is larger, the corresponding cosine adjustment factor will be adjusted closer to 1 for larger amplitude adjustments in the subsequent optimization process. Conversely, for those parameters for which the target variance data is smaller, the remaining string adjustment factors will be adjusted closer to 0 to maintain the current setting.
In a specific embodiment, the process of executing step 103 may specifically include the following steps:
(1) Carrying out chaotic mapping on the cosine adjustment factors through a preset Logistic chaotic mapping function to generate nonlinear oscillation adjustment factors of each first image quality processing parameter in the initialization parameter population;
(2) Determining a parameter search space and iteration times of an initialization parameter population according to the nonlinear oscillation adjustment factor;
(3) And carrying out parameter searching on the initialized parameter population according to the parameter searching space and the iteration times to generate a plurality of second image quality processing parameters.
Specifically, the cosine adjustment factor is subjected to chaotic mapping through a preset Logistic chaotic mapping function, a series of very complex and unpredictable output results are generated, and the output results serve as nonlinear oscillation adjustment factors and reflect the potential direction and amplitude for adjusting each first image quality processing parameter. The parameter search space and the iteration times of the initialized parameter population are determined according to the nonlinear oscillation adjustment factors. The determination of the parameter search space is based on the range of variation of the adjustment potential sum of each parameter. For example, if a nonlinear oscillation adjustment factor of a certain parameter indicates that it requires a large range of adjustment, its parameter search space will be set wider, and conversely narrower. Meanwhile, the determination of the iteration times needs to consider the complexity of the optimization process and the requirement of solving precision, and can be ensured to be effectively converged when the sufficient precision is achieved. And carrying out parameter searching on the initialized parameter population according to the determined parameter searching space and iteration times. In each iteration, parameters are adjusted according to the direction and the amplitude indicated by the nonlinear oscillation adjustment factor, and then the influence of the adjusted parameters on the image quality is evaluated. The chaotic mapping function-based method not only can improve the randomness and coverage of searching, but also can help the algorithm jump out of a local optimal solution, and more potential optimization directions are explored. For example, assume that a parameter in the initialized parameter population generates an adjustment factor for large-amplitude oscillations via chaotic mapping, which indicates that significant adjustments to the parameter are required during the optimization process. Thus, during the parameter search, the parameter will be iteratively tested in a larger search space to find the best parameter setting. After a plurality of iterations, the finally generated second image quality processing parameters better promote the texture details of the digital character image.
In a specific embodiment, the process of executing step 104 may specifically include the following steps:
(1) Inputting each second image quality processing parameter into a preset image quality processing parameter prediction model, wherein the image quality processing parameter prediction model comprises: a convolution long-short time memory network, a threshold circulation network and a linear regression layer;
(2) Performing parameter feature space mapping on each second image quality processing parameter through a convolution long-short-term memory network to obtain a corresponding parameter feature vector;
(3) Inputting the parameter feature vector into a threshold cyclic network for feature coding to obtain a parameter coding vector, inputting the parameter coding vector into a linear regression layer for image quality rendering simulation operation to obtain target image quality rendering parameters of each second image quality processing parameter;
(4) The method comprises the steps of performing multi-modal parameter classification on target image quality rendering parameters to obtain a plurality of digital person modal data, and performing feature extraction on the plurality of digital person modal data to obtain a plurality of digital person modal features;
(5) And carrying out feature fusion on the plurality of digital human modal features to obtain target multi-modal fusion features of each second image quality processing parameter.
Specifically, each second image quality processing parameter is input into a preset image quality processing parameter prediction model, and the model comprehensively utilizes a convolution long-short time memory network (ConvLSTM), a threshold cycle network (GRU) and a linear regression layer. And performing parameter feature space mapping on each second image quality processing parameter through a convolution long-short-term memory network. ConvLSTM is a specially designed neural network that combines the ability of Convolutional Neural Networks (CNNs) to process spatial information with the ability of long-term memory networks (LSTM) to process time-series data. Through this network, each image quality processing parameter is mapped to a parameter feature space, generating a corresponding parameter feature vector. This vector contains not only the original information of the parameters, but also incorporates the complex nature of the parameters in both spatial and temporal dimensions. These parameter feature vectors are input to a threshold cycle network. The GRU is a highly efficient Recurrent Neural Network (RNN) that controls the flow of information through specific gantry mechanisms, which allows it to maintain long-term dependencies while processing sequence data, and avoid gradient vanishing problems. In this link, the GRU performs feature encoding on the input parameter feature vector to generate a parameter encoding vector. These encoded vectors are further abstractions and refinements to the original feature vectors, which capture the most core and critical information about the image quality processing parameters. These parameter code vectors are input to the linear regression layer. The linear regression layer converts the complex coded vectors into the output of a specific image quality rendering simulation operation. Through this layer of processing, target image quality rendering parameters for each of the second image quality processing parameters are generated, which are optimizations and improvements to the original second image quality processing parameters that predict how the parameters affect the final image quality. And then, carrying out multi-modal parameter classification on the target image quality rendering parameters to generate a plurality of digital human modal data. The image quality rendering parameters are classified according to different visual and technical characteristics, such as texture, color, brightness and the like. Feature extraction is performed on the classified modal data to generate a plurality of digital human modal features. These features represent the specifics and effects of the parameters in each modality. And combining the digital human features of different modes through a feature fusion technology to obtain target multi-mode fusion features of each second image quality processing parameter. The purpose of feature fusion is to integrate information from different modalities to provide a more comprehensive and detailed view to evaluate and predict the effect of image quality processing parameters.
In a specific embodiment, the process of executing step 105 may specifically include the following steps:
(1) Inputting the target multi-mode fusion characteristics into a preset cascading forest model, wherein the cascading forest model comprises a plurality of decision trees;
(2) Image quality processing decision analysis is respectively carried out on the target multi-mode fusion characteristics through a plurality of decision trees, so that an initial decision evaluation index of each decision tree is obtained;
(3) And obtaining weight data corresponding to the decision trees, and carrying out evaluation index weighted analysis on the initial decision evaluation index of each decision tree according to the weight data to obtain image quality rendering characteristic data of each second image quality processing parameter.
Specifically, the target multi-mode fusion characteristic is input into a preset cascade forest model. The characteristics comprise information such as colors, textures, edges and the like of the images, and each decision tree in the cascade forest model independently analyzes the characteristics and outputs own decision evaluation indexes. These indices represent the judgment and evaluation of the image quality processing parameters for each tree. And obtaining weight data corresponding to each decision tree. These weight data are calculated from the historical performance and accuracy of each tree. The tree with better effect and more accurate judgment is given higher weight, and the output initial decision evaluation index of the tree occupies larger proportion in the final comprehensive analysis. For example, if a tree exhibits higher accuracy and reliability in past image quality processing tasks, its judgment in the current task is given higher weight. The initial decision evaluation index of each tree will be weighted according to the corresponding weight. The intelligence and the judgment of all decision trees are integrated, and more reliable and effective image quality processing decisions are ensured by a weighting mode. For example, if some trees consider increasing contrast as a key to improving image quality and the weight of these trees is higher, then the suggestion to increase contrast will be more important in the final decision. By this weighted analysis method, image quality rendering characteristic data of each second image quality processing parameter is obtained. The data reflects how to adjust the image quality processing parameters to achieve the best rendering effect under the comprehensive consideration of various factors and decision trees. For example, when processing a night scene image, by cascading analysis of the forest model, it is concluded that properly increasing brightness and saturation, while reducing noise, can make the image clearer and more vivid.
In a specific embodiment, the process of executing step 106 may specifically include the following steps:
(1) Comparing the image quality rendering characteristic data with a first evaluation index threshold and a second evaluation index threshold, wherein the first evaluation index threshold is smaller than the second evaluation index threshold;
(2) If the image quality rendering characteristic data is less than the first evaluation index threshold value, dividing the corresponding second image quality processing parameter into a first parameter population, if the first evaluation index threshold value is less than the image quality rendering characteristic data is less than the second evaluation index threshold value, dividing the corresponding second image quality processing parameter into a second parameter population, and if the second evaluation index threshold value is less than the image quality rendering characteristic data, dividing the corresponding second image quality processing parameter into a third parameter population;
(3) Propagating, intersecting and mutating the first parameter population and the second parameter population, and intersecting and mutating the third parameter population to obtain a plurality of candidate image quality processing parameters;
(4) And carrying out optimization analysis on the plurality of candidate image quality processing parameters to obtain target image quality processing parameters.
Specifically, two evaluation index thresholds are set: a first evaluation index threshold and a second evaluation index threshold, wherein the first evaluation index threshold is lower than the second evaluation index threshold. The setting of these two thresholds is based on the desire and requirement for image quality processing effects, which will be the criteria for dividing the different quality level image quality processing parameters. The image quality rendering characteristic data is compared and analyzed to determine which parameter population each image quality processing parameter belongs to. If the image quality rendering characteristic data is smaller than the first evaluation index threshold, this indicates that the parameter has a smaller effect on improvement of the image quality, and is therefore classified into the first parameter population. This population contains relatively weak image quality processing parameters, requiring further optimization or improvement. For example, a parameter that slightly increases the brightness of the image may be classified into the population if it does not improve sufficiently to reach the first evaluation index threshold. For those parameters of the image quality rendering characteristic data that are between the first and second evaluation index thresholds, they are divided into a second parameter population. The parameters in this population are relatively good, they have a significant effect on the improvement of the image quality, but have not yet reached an optimal state. For example, a parameter that performs well in detail processing may be classified as a second parameter population if its effect exceeds a first evaluation index threshold but does not reach a second evaluation index threshold. If the image quality rendering characteristic data exceeds the second evaluation index threshold, the parameters are divided into a third parameter population. The parameters of this population have met or exceeded the set high standards in image quality processing. For example, a parameter that significantly improves the color saturation and contrast of an image may be classified as a third parameter if its effect exceeds the second evaluation index threshold. The breeding, crossover and mutation operations are performed on the first and second parameter populations. Parameters are optimized by modeling natural selection and genetic mechanisms. Reproduction refers to the selection of parameters that perform well for replication, crossover refers to the combination of two parameters to produce new parameters, and mutation refers to the random minor modification of the parameters. And for the third parameter population, the crossover and mutation operations are directly performed. Since the parameters in this population are very good themselves, no propagation is required, and it is directly through crossover and mutation that explores whether there is room for further optimization. Through the steps, a plurality of candidate image quality processing parameters are obtained. And carrying out optimization analysis on the candidate parameters, carrying out comprehensive evaluation on the candidate parameters, and examining the effect of the candidate parameters in practical application. This evaluation is based on various factors such as sharpness of image quality, authenticity of color, degree of retention of details, and the like. Through the process, parameters which can truly improve the image quality are screened out, and parameters which are good in theory but have general effects in practical application are excluded.
The above describes the image quality processing method of the metauniverse digital person in the embodiment of the present application, and the following describes the image quality processing device of the metauniverse digital person in the embodiment of the present application, referring to fig. 2, one embodiment of the image quality processing device of the metauniverse digital person in the embodiment of the present application includes:
the initialization module 201 is configured to perform population initialization on image quality processing parameters of a target meta-universe digital person by adopting a sine and cosine algorithm to obtain an initialization parameter population, where the initialization parameter population includes: a plurality of first image quality processing parameters;
A calculation module 202, configured to calculate target fitness data of each first image quality processing parameter, and dynamically adjust a cosine adjustment factor according to the target fitness data;
The construction module 203 is configured to construct a nonlinear oscillation adjustment factor of the initialization parameter population according to the cosine adjustment factor, and perform parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factor, so as to generate a plurality of second image quality processing parameters;
the extracting module 204 is configured to perform image quality rendering simulation operation on the target meta-universe digital person according to the plurality of second image quality processing parameters, obtain a target image quality rendering parameter of each second image quality processing parameter, and perform multi-modal fusion feature extraction on the target image quality rendering parameter, so as to obtain a target multi-modal fusion feature of each second image quality processing parameter;
the analysis module 205 is configured to input the target multi-mode fusion feature into a preset cascade forest model to perform image quality rendering feature analysis, so as to obtain image quality rendering feature data of each second image quality processing parameter;
And an optimization module 206, configured to perform an optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering feature data, so as to obtain target image quality processing parameters.
Through the cooperation of the components, the multilevel optimization of the initialization parameter population is realized through the dynamic adjustment of a sine and cosine algorithm and a cosine adjustment factor, and the breadth and the depth of the search space are improved. And the nonlinear oscillation adjustment factor is utilized for parameter searching, so that the flexibility of an algorithm is effectively increased, and the characteristic of image quality processing parameters can be more comprehensively captured. The multi-mode fusion feature extraction is carried out by adopting a convolution long-short-time memory network, a threshold circulation network and a linear regression layer, so that the multi-mode fusion feature extraction is beneficial to comprehensively considering the image quality information in many aspects such as texture, depth scene, color saturation and the like. The feature extraction mode enables the system to be more suitable for the complex image quality requirements of metauniverse digital people, and improves the global perception capability of image quality rendering. And a cascading forest model is introduced to conduct image quality rendering feature analysis, and decision analysis is conducted through a plurality of decision trees, so that different image quality processing situations can be understood and processed more carefully. The use of the model improves the decision accuracy and the robustness of the system, so that the image quality processing is more intelligent and adaptive. And a mechanism for dynamically adjusting the cosine adjustment factor is adopted, and real-time adjustment is carried out according to the target fitness data, so that the adaptability of the algorithm to different image quality scenes is improved. The adaptivity enables the system to be more robust when processing different digital human image quality, and can cope with diversified metauniverse scenes and demands. The searching and selecting efficiency of the target image quality processing parameters is improved through comprehensive optimization strategies such as division, propagation, crossing and variation of a plurality of parameter populations. The comprehensive optimization strategy is beneficial to finding out better image quality processing parameters in a large-scale parameter space, so that the overall image quality rendering effect is improved, and the image quality processing accuracy of the metauniverse digital person is further improved.
The present application also provides a computer device including a memory and a processor, the memory storing computer readable instructions that, when executed by the processor, cause the processor to perform the steps of the method for processing image quality of a metauniverse digital person in the above embodiments.
The present application also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, or may be a volatile computer readable storage medium, in which instructions are stored, which when executed on a computer, cause the computer to perform the steps of the method for processing image quality of a metadigital person.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a read-oXly memory (ROM), a random access memory (raXdom access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims (10)

1. The image quality processing method of the metauniverse digital person is characterized by comprising the following steps of:
Carrying out population initialization on image quality processing parameters of a target element universe digital person by adopting a sine and cosine algorithm to obtain an initialization parameter population, wherein the initialization parameter population comprises: a plurality of first image quality processing parameters;
Respectively calculating target fitness data of each first image quality processing parameter, and dynamically adjusting a cosine adjustment factor according to the target fitness data;
Constructing a nonlinear oscillation adjustment factor of the initialization parameter population according to the cosine adjustment factor, and carrying out parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factor to generate a plurality of second image quality processing parameters;
Performing image quality rendering simulation operation on the target meta-universe digital person according to the second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter;
Inputting the target multi-mode fusion characteristics into a preset cascade forest model for image quality rendering characteristic analysis to obtain image quality rendering characteristic data of each second image quality processing parameter;
And carrying out optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering characteristic data to obtain target image quality processing parameters.
2. The method for processing the image quality of the metauniverse digital person according to claim 1, wherein the method for processing the image quality of the target metauniverse digital person by using a sine and cosine algorithm is characterized in that a population is initialized to obtain an initialization parameter population, wherein the initialization parameter population comprises: a plurality of first image quality processing parameters, comprising:
Defining an image quality processing parameter space, wherein the image quality processing parameter space comprises a plurality of image quality processing parameters and parameter intervals corresponding to each image quality processing parameter, and the plurality of image quality processing parameters comprise: texture map parameters, depth scene parameters, and color saturation parameters;
Generating uniformly distributed random values for each image quality processing parameter in the image quality processing parameter space through a sine and cosine algorithm;
Generating an initialization parameter population of the target metauniverse digital person according to the random value of each image quality processing parameter, wherein the initialization parameter population comprises the following components: a plurality of first image quality processing parameters.
3. The method for processing the image quality of the metauniverse digital person according to claim 1, wherein the calculating the target fitness data of each first image quality processing parameter, respectively, and dynamically adjusting the cosine adjustment factor according to the target fitness data, comprises:
Determining a value range of a cosine adjusting factor, wherein the value range is [0,1], 0 represents no adjustment, and 1 represents the maximum adjustment;
Respectively calculating target fitness data of each first image quality processing parameter, and acquiring preset standard fitness data;
Calculating difference data between the target fitness data and the standard fitness data to obtain target difference data of each first image quality processing parameter;
and calculating average difference data corresponding to the target difference data of each first image quality processing parameter, and dynamically adjusting cosine adjustment factors of each first image quality processing parameter according to the average difference data and the value range.
4. The method of claim 3, wherein constructing a nonlinear oscillation adjustment factor of the initialization parameter population according to the cosine adjustment factor, and performing a parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factor, and generating a plurality of second image quality processing parameters, comprises:
Performing chaotic mapping on the cosine adjustment factors through a preset Logistic chaotic mapping function to generate nonlinear oscillation adjustment factors of each first image quality processing parameter in the initialization parameter population;
Determining a parameter search space and iteration times of the initialization parameter population according to the nonlinear oscillation adjustment factor;
And carrying out parameter searching on the initialized parameter population according to the parameter searching space and the iteration times to generate a plurality of second image quality processing parameters.
5. The method according to claim 1, wherein performing an image quality rendering simulation operation on the target metauniverse digital person according to the plurality of second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter, comprises:
inputting each second image quality processing parameter into a preset image quality processing parameter prediction model, wherein the image quality processing parameter prediction model comprises: a convolution long-short time memory network, a threshold circulation network and a linear regression layer;
Performing parameter feature space mapping on each second image quality processing parameter through the convolution long-short time memory network to obtain a corresponding parameter feature vector;
inputting the parameter feature vector into the threshold cyclic network for feature coding to obtain a parameter coding vector, and inputting the parameter coding vector into the linear regression layer for image quality rendering simulation operation to obtain target image quality rendering parameters of each second image quality processing parameter;
The target image quality rendering parameters are subjected to multi-modal parameter classification to obtain a plurality of digital human modal data, and feature extraction is performed on the plurality of digital human modal data to obtain a plurality of digital human modal features;
and carrying out feature fusion on the digital human modal features to obtain target multi-modal fusion features of each second image quality processing parameter.
6. The method for processing the image quality of the metauniverse digital person according to claim 1, wherein inputting the target multi-mode fusion feature into a preset cascade forest model for image quality rendering feature analysis, obtaining image quality rendering feature data of each second image quality processing parameter, comprises:
inputting the target multi-mode fusion characteristics into a preset cascading forest model, wherein the cascading forest model comprises a plurality of decision trees;
image quality processing decision analysis is respectively carried out on the target multi-mode fusion characteristics through the decision trees, so that an initial decision evaluation index of each decision tree is obtained;
and acquiring weight data corresponding to the decision trees, and carrying out evaluation index weighted analysis on the initial decision evaluation index of each decision tree according to the weight data to obtain image quality rendering characteristic data of each second image quality processing parameter.
7. The method according to claim 1, wherein the optimizing the plurality of second image quality processing parameters based on the image quality rendering characteristic data to obtain the target image quality processing parameters comprises:
Comparing the image quality rendering characteristic data with a first evaluation index threshold and a second evaluation index threshold, wherein the first evaluation index threshold is smaller than the second evaluation index threshold;
If the image quality rendering characteristic data is less than the first evaluation index threshold value, dividing the corresponding second image quality processing parameter into a first parameter population, if the first evaluation index threshold value is less than the image quality rendering characteristic data is less than the second evaluation index threshold value, dividing the corresponding second image quality processing parameter into a second parameter population, and if the second evaluation index threshold value is less than the image quality rendering characteristic data, dividing the corresponding second image quality processing parameter into a third parameter population;
Reproducing, intersecting and mutating the first parameter population and the second parameter population, and intersecting and mutating the third parameter population to obtain a plurality of candidate image quality processing parameters;
And carrying out optimization analysis on the plurality of candidate image quality processing parameters to obtain target image quality processing parameters.
8. An image quality processing apparatus for a metauniverse digital person, the image quality processing apparatus comprising:
the initialization module is used for carrying out population initialization on image quality processing parameters of the target universe digital person by adopting a sine and cosine algorithm to obtain an initialization parameter population, wherein the initialization parameter population comprises: a plurality of first image quality processing parameters;
The computing module is used for respectively computing target fitness data of each first image quality processing parameter and dynamically adjusting cosine adjustment factors according to the target fitness data;
The construction module is used for constructing nonlinear oscillation adjustment factors of the initialization parameter population according to the cosine adjustment factors, and carrying out parameter search on the initialization parameter population according to the nonlinear oscillation adjustment factors to generate a plurality of second image quality processing parameters;
The extraction module is used for performing image quality rendering simulation operation on the target meta-universe digital person according to the plurality of second image quality processing parameters to obtain target image quality rendering parameters of each second image quality processing parameter, and performing multi-modal fusion feature extraction on the target image quality rendering parameters to obtain target multi-modal fusion features of each second image quality processing parameter;
the analysis module is used for inputting the target multi-mode fusion characteristics into a preset cascade forest model to perform image quality rendering characteristic analysis to obtain image quality rendering characteristic data of each second image quality processing parameter;
and the optimization module is used for carrying out optimization analysis on the plurality of second image quality processing parameters according to the image quality rendering characteristic data to obtain target image quality processing parameters.
9. A computer device, the computer device comprising: a memory and at least one processor, the memory having instructions stored therein;
The at least one processor invoking the instructions in the memory to cause the computer device to perform the method of image quality processing of a metauniverse digital person as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium having instructions stored thereon, which when executed by a processor, implement the method of image quality processing of a metauniverse digital person according to any one of claims 1 to 7.
CN202410008023.0A 2024-01-04 2024-01-04 Image quality processing method, device, equipment and storage medium for metauniverse digital person Active CN117523060B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410008023.0A CN117523060B (en) 2024-01-04 2024-01-04 Image quality processing method, device, equipment and storage medium for metauniverse digital person

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410008023.0A CN117523060B (en) 2024-01-04 2024-01-04 Image quality processing method, device, equipment and storage medium for metauniverse digital person

Publications (2)

Publication Number Publication Date
CN117523060A CN117523060A (en) 2024-02-06
CN117523060B true CN117523060B (en) 2024-05-17

Family

ID=89744224

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410008023.0A Active CN117523060B (en) 2024-01-04 2024-01-04 Image quality processing method, device, equipment and storage medium for metauniverse digital person

Country Status (1)

Country Link
CN (1) CN117523060B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111985599A (en) * 2020-07-31 2020-11-24 西安理工大学 Improved sine and cosine algorithm
CN112085059A (en) * 2020-08-06 2020-12-15 温州大学 Breast cancer image feature selection method based on improved sine and cosine optimization algorithm
CN115330617A (en) * 2022-08-11 2022-11-11 潘奕旻 Medical image self-adaptive enhancement method based on improved multivariate universe optimization algorithm
CN116776254A (en) * 2023-06-30 2023-09-19 江西理工大学 Spark and NRSCA strategy parallel depth forest algorithm for regional landslide hazard evaluation
CN117011130A (en) * 2022-09-20 2023-11-07 腾讯科技（深圳）有限公司 Method, apparatus, device, medium and program product for image super resolution

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230316594A1 (en) * 2022-03-29 2023-10-05 Meta Platforms Technologies, Llc Interaction initiation by a virtual assistant

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111985599A (en) * 2020-07-31 2020-11-24 西安理工大学 Improved sine and cosine algorithm
CN112085059A (en) * 2020-08-06 2020-12-15 温州大学 Breast cancer image feature selection method based on improved sine and cosine optimization algorithm
CN115330617A (en) * 2022-08-11 2022-11-11 潘奕旻 Medical image self-adaptive enhancement method based on improved multivariate universe optimization algorithm
CN117011130A (en) * 2022-09-20 2023-11-07 腾讯科技（深圳）有限公司 Method, apparatus, device, medium and program product for image super resolution
CN116776254A (en) * 2023-06-30 2023-09-19 江西理工大学 Spark and NRSCA strategy parallel depth forest algorithm for regional landslide hazard evaluation

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Yaming Wang et al..Image super-resolution based on multi-grained cascade forests.《International Journal of Wavelets,Multiresolution and Information Processing》.2019,第17卷(第4期),第1-14页. *
基于Logistic模型和随机差分变异的正弦余弦算法;徐明 等;《计算机科学》;20201231(第02期);第212-218页 *

Also Published As

Publication number Publication date
CN117523060A (en) 2024-02-06

Similar Documents

Publication Publication Date Title
Sitthi-Amorn et al. Genetic programming for shader simplification
Zhao et al. Achieving good connectivity in motion graphs
CN109165720A (en) Neural network model compression method, device and computer equipment
CN109829959B (en) Facial analysis-based expression editing method and device
Ulhaq et al. Efficient diffusion models for vision: A survey
TWI655587B (en) Neural network and method of neural network training
CN112418292B (en) Image quality evaluation method, device, computer equipment and storage medium
CN113723530B (en) Intelligent psychological assessment system based on video analysis and electronic psychological sand table
CA3137297C (en) Adaptive convolutions in neural networks
CN110210204B (en) Verification code generation method and device, storage medium and electronic equipment
WO2021144943A1 (en) Control method, information processing device, and control program
Moua et al. Correcting the effect of sampling bias in species distribution modeling–A new method in the case of a low number of presence data
Dhou et al. An innovative chain coding technique for compression based on the concept of biological reproduction: an agent-based modeling approach
Cavallaro et al. Machine Learning and Genetic Algorithms: A case study on image reconstruction
CN117523060B (en) Image quality processing method, device, equipment and storage medium for metauniverse digital person
US11763579B2 (en) Obtaining patterns for surfaces of objects
Cui et al. Single bitmap block truncation coding of color images using cat swarm optimization
CN115879507A (en) Large-scale graph generation method based on deep confrontation learning
Li et al. Aesthetic learning in an interactive evolutionary art system
Saini et al. Image compression using APSO
CN112439200B (en) Data processing method, data processing device, storage medium and electronic equipment
CN117376538B (en) Response speed optimization method, device, equipment and medium of laser projection optical machine
CN111178541B (en) Game artificial intelligence system and performance improving system and method thereof
CN116704588B (en) Face image replacing method, device, equipment and storage medium
Bader et al. SID-avatar database: A 3D Avatar Dataset for virtual world research

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant