Neural Network Auxiliary Input Positioning for Adaptive Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression technologies face challenges in achieving high compression ratios with minimal quality loss, particularly in limited bandwidth scenarios, and there is a need for improved configurability and adaptability in neural network-based encoding and decoding processes.
Innovation Solution
A method and apparatus that allow for configurable positions within a neural network to gather auxiliary data based on preconfigured conditions, enabling dynamic architecture changes and flexible data processing, including the use of prediction data and supplementary data at specific layers to enhance encoding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If video data is compressed using traditional methods, then the data size is reduced for transmission, but the picture quality deteriorates
Solution Approach 1:
The patent replaces traditional mechanical video compression algorithms with a neural network-based system. The neural network learns optimal compression transformations during training, substituting conventional block-based DCT transforms and quantization methods with learned representations that better preserve visual quality while achieving compression.
Solution Approach 2:
The patent changes the fundamental parameters of video compression by using end-to-end learned transformations instead of fixed mathematical transforms. The neural network adjusts compression parameters dynamically based on content characteristics, replacing static quantization matrices and block sizes with adaptive learned parameters.
2Ease of manufacture
If a fixed neural network architecture is used for video processing, then the implementation is simple, but the adaptability to different content characteristics is poor
Solution Approach 1:
The patent introduces dynamic elements into the neural network architecture by allowing the selection of different processing paths based on content characteristics. The system dynamically adjusts which neural network components are activated and how auxiliary information is gathered and processed, enabling adaptation to different video content types while maintaining a relatively simple base architecture.
Solution Approach 2:
The patent creates a universal neural network framework that can handle multiple video processing tasks and content types through a single architecture. The system uses configurable auxiliary information gathering at multiple positions to serve different content characteristics, making the same base network applicable to various video scenarios without requiring separate specialized networks.
3Measurement precision
If auxiliary information is gathered at all preconfigured positions in the neural network, then the processing accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies local quality by selectively gathering auxiliary information only at specific preconfigured positions within the neural network where it provides the most benefit. Different regions of the network receive auxiliary information based on their specific processing needs, rather than uniformly applying it throughout, thereby improving accuracy where needed while limiting complexity growth.
Solution Approach 2:
The patent uses partial action by gathering auxiliary information at a subset of preconfigured positions rather than all positions. The system selectively applies auxiliary information gathering based on configuration parameters, implementing partial coverage that provides sufficient processing accuracy improvement without the full computational overhead of gathering at every possible position.
Data Source
AI summary
Methods and apparatuses for processing of picture data or picture feature data using a neural network with two or more layers are provided. The present disclosure may be applied in the field of artificial intelligence (AI)-based video or picture compression technologies, and in particular, to the field of neural network-based video compression technologies. According to the present disclosure, position within the neural network, at which auxiliary information may be entered for processing is selectable based on a gathering condition. The gathering condition may assess whether some prerequisite is fulfilled. Advantages may include better performance in terms of rate and/or disclosure due to the effect of increased flexibility in neural network configurability.