Feature Channel Ordering for Higher Video Encoding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding methods for feature data suffer from low encoding efficiency due to the lack of guidance on the reference relationship among feature channels, leading to weak correlation among data channels and inefficient use of existing video coding technologies.
Innovation Solution
A method for encoding feature data involves sorting channels by similarity and splicing them into a target frame sequence, followed by encoding using optimized inter-reference structures, and decoding the reconstructed frames for further analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If feature maps are superimposed in a specified order using a single list, then the feature channels can be organized in a defined sequence, but the encoding efficiency of the feature channels is not high due to lack of guidance on reference relationships
Solution Approach 1:
The patent changes the ordering parameter of feature channels from a simple single list to a hierarchical structure with multiple lists (first list and second list). This parameter change enables different reference relationships to be established, where the first list provides primary reference and the second list provides additional reference, thereby improving encoding efficiency while maintaining organized structure.
2Stability of the object's composition
If multi-channel feature data are tiled in a picture according to a single list order, then the multi-channel data are closely adjacent, but the correlation among data of different channels in the same coding unit is weak due to discontinuity
Solution Approach 1:
The patent segments the feature channel organization into multiple lists (first list and second list) rather than using a single flat list. This segmentation allows different groups of channels to be organized with different reference relationships, enabling channels with stronger correlations to be placed in the same coding unit while maintaining overall adjacency of multi-channel data.
3Device complexity
If existing methods for processing feature data are used, then the process is simple, but the efficiency cannot be effectively exerted due to weak correlation among channels
Solution Approach 1:
The patent introduces dynamic adaptability by allowing different reference relationships (first list and second list) to be selected based on the specific characteristics of the feature data. This dynamic approach enables the system to adapt to different data correlation patterns, improving processing efficiency without requiring overly complex fixed structures.
Data Source
AI summary
Embodiments of the disclosure provide a method for encoding feature data, a method for decoding feature data, and a decoder. The method for encoding feature data includes the following. Feature data of multiple channels corresponding to a picture to-be-processed is obtained. Feature data of a reference channel in the feature data of the multiple channels is determined. Sorted feature data of the multiple channels is obtained by sorting, starting from the feature data of the reference channel, the feature data of the multiple channels in a descending order of similarity among the feature data of the multiple channels. The sorted feature data of the multiple channels is spliced to obtain a target feature frame sequence. The target feature frame sequence is encoded to generate a bitstream.