Bitstream Index Compression for Parallel Entropy Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-quality video data places a significant burden on communication networks and devices due to the large amount of data required, and existing video coding techniques face throughput bottlenecks and compression losses when employing parallelization and machine learning-based methods.
Innovation Solution
The implementation of systems and techniques for generating and decoding bitstream indexes in parallel entropy coding, using neural networks to identify entry points for entropy codable parcels and reduce overhead, thereby enhancing compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallel entropy coding is implemented to increase throughput, then processing speed is improved, but compression efficiency deteriorates due to overhead and bottlenecks
Solution Approach 1:
The patent segments the bitstream into independently entropy-codable parcels, allowing parallel processing of multiple parcels simultaneously. This segmentation enables throughput improvement while maintaining compression efficiency by avoiding the bottlenecks of sequential processing.
Solution Approach 2:
The patent introduces a new dimension of parallelization by processing multiple entropy-codable parcels concurrently rather than sequentially. This dimensional change from single-threaded to multi-threaded processing resolves the throughput bottleneck while the index structure maintains compression efficiency.
2Manufacturing precision
If machine learning-based methods are used to improve coding efficiency, then compression performance is improved, but computational complexity increases
Solution Approach 1:
The patent divides the complex machine learning-based encoding task into smaller, independently processable parcels. Each parcel can be processed by separate neural network instances in parallel, reducing the computational complexity of individual processing units while maintaining overall coding efficiency.
Solution Approach 2:
The patent performs preliminary identification of entry points and parcel boundaries before the main entropy coding process. This preliminary action allows the neural network to be configured optimally for parallel processing, reducing computational complexity during the actual encoding phase while preserving coding efficiency.
3Productivity
If the bitstream is divided into multiple parcels for parallel processing, then throughput is improved, but overhead increases due to index generation
Solution Approach 1:
The patent generates an index only for the necessary entry points of parcels rather than for every possible data element. This partial action approach provides sufficient information for parallel decoding while minimizing the overhead associated with index generation and storage.
Solution Approach 2:
The patent optimizes the granularity and structure of the parcel index to balance throughput improvement against overhead increase. By carefully controlling the number and positioning of entry points, the system achieves parallel processing benefits while keeping the index size and associated overhead manageable.
Data Source
AI summary
Systems and techniques are described herein for processing video data. For example, an encoding device can obtain a sequence of video data and determine a minimum value in the sequence of video data. The encoding device can, based on the minimum value, identify positions in the sequence of video data associated with entry points for individually entropy codable parcels of a parallel entropy codable sequence of video data. The encoding device can generate the parallel entropy codable sequence of video data. The encoding device can further generate an index for the parallel entropy codable sequence of video data, the index identifying the individually entropy codable parcels within the parallel entropy codable sequence of video data.