Neural Video Encoding with Adaptive Encoder-Decoder Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression and reconstruction technologies using neural network models face challenges in achieving efficient data compression and transmission while adhering to footprint constraints, particularly in systems-on-chip (SoC) environments, where power consumption and bandwidth management are critical.
Innovation Solution
A data processing method utilizing a system comprising multiple encoders and decoders trained on different loss functions, where a neural network model selects optimal encoder-decoder pairs for each image block region, generating encoded data with identification data for decoding, and employing entropy encoding to optimize compression and transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple encoders and decoders are used to improve compression performance, then data compression efficiency is improved, but device complexity increases
Solution Approach 1:
The system divides the encoding task into multiple independent encoder units, each processing different aspects or regions of the input data. This segmentation allows parallel processing and improves overall compression efficiency while maintaining manageable complexity through modular architecture
Solution Approach 2:
The patent implements dynamic selection mechanisms that adaptively choose which encoders and decoders to activate based on input data characteristics and footprint constraints. This dynamic approach optimizes compression performance for each specific task while managing system resources, effectively resolving the contradiction between performance and complexity
2Productivity
If neural network models are used for feature learning to improve compression ratio, then data compression performance is improved, but power consumption increases
Solution Approach 1:
The system applies partial action by selectively activating only the necessary encoder-decoder pairs based on footprint constraints and input data requirements. Rather than running all neural network models continuously, the system activates only the required subset, reducing power consumption while maintaining high compression ratio for the given task constraints
Solution Approach 2:
The patent changes operational parameters by adjusting the activation state of different encoder-decoder pairs based on footprint constraints. This parameter adjustment allows the system to optimize between compression performance and power consumption by selecting appropriate model configurations for different operating conditions
3Quantity of substance
If entropy encoding is applied to increase data compression efficiency, then transmission bandwidth is reduced, but processing time increases
Solution Approach 1:
The system performs preliminary feature extraction and encoding using neural network models before applying entropy encoding. This preliminary action prepares the data in a way that reduces the complexity of subsequent entropy encoding operations, achieving efficient compression with lower processing time overhead
Data Source
AI summary
A data processing method may comprise receiving input data, obtaining a plurality of vector values from a plurality of encoders by inputting the input data to the plurality of encoders, selecting a vector values to be included in encoded data from the plurality of vector values using a neural network model that receives the vector values as input, and generating the encoded data comprising the selected vector values and identification data that identifies a decoder to decode the selected vector values among a plurality of decoders.