Adaptive Neural Image Coding for Device-Specific Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network-based picture encoding and decoding schemes have a fixed network structure, failing to meet the diverse requirements of different application scenarios, particularly in terms of computing power and compression efficiency across devices with varying capabilities.
Innovation Solution
A method and apparatus that dynamically adjust the encoder and decoder networks based on profile information, allowing selection of different network structures to balance delay and compression performance, enabling flexibility in decoding performance across devices with varying computing power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a fixed neural network structure is used for picture encoding and decoding, then the system is simple to implement, but it cannot meet the diverse requirements of different application scenarios with varying computing power and compression efficiency needs
Solution Approach 1:
The patent implements dynamic network structure selection by introducing multiple decoder networks with different complexities (first decoder network with higher processing resources, second decoder network with lower processing resources) and using identification information in the bitstream to dynamically select which network to use. This allows the system to adapt its complexity to match the terminal device's computing power and the specific application scenario requirements.
Solution Approach 2:
The patent changes the parameter of network structure by encoding identification information that indicates different decoder network types. The terminal device decodes this identification information and selects the appropriate decoder network accordingly. This parameter change approach enables the system to switch between different network configurations without requiring physical hardware changes, thus achieving adaptability while maintaining manageable system complexity.
2Manufacturing precision
If a high-processing-resource decoder network is used, then decoding quality and compression efficiency are improved, but the processing resource consumption increases
Solution Approach 1:
The patent enables dynamic adjustment of processing resource consumption by changing the network structure parameter based on the identification information decoded from the bitstream. When the identified scenario requires high decoding quality, the first decoder network with more processing resources is selected. When lower quality is acceptable or computing power is limited, the second decoder network with fewer processing resources is selected. This parameter-based selection mechanism allows flexible balancing between decoding quality and processing resource consumption.
3Manufacturing precision
If a high-processing-resource encoder network is used, then compression performance is improved, but the encoding delay increases
Solution Approach 1:
The patent implements dynamic encoder network selection by using identification information to determine which encoder network (first with higher processing resources, second with lower processing resources) to use for encoding. This dynamic selection allows the system to optimize the balance between compression performance and encoding delay based on the specific application scenario and terminal device capabilities, rather than being constrained by a fixed encoder configuration.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A picture encoding and decoding method and apparatus are provided, and relate to the artificial intelligence field and the picture compression field, to provide an encoding and decoding scheme, thereby meeting requirements of different application scenarios. According to the encoding and decoding method provided in this application, a used encoder and decoder network may be determined based on profile information (or identification information). That is, a codec may select corresponding profile information based on a capability of a decoding device, to select or indicate different encoder and decoder networks. In this way, the network may not only have a capability of adapting to a terminal side with low computing power, but also have a capability of adapting to a terminal side with higher computing power.