Fast Multi-Rate HEVC Encoding for Adaptive Streaming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-quality video streaming on mobile devices with varying bandwidths strains storage and encoding resources, as existing HEVC encoders face high computational complexity in generating multiple representations, particularly in live DASH streaming, due to the need for optimal CU depth structure and PU mode decisions across different bit-rates and resolutions.
Innovation Solution
A method for fast multi-rate encoding that exploits redundancy between co-located CTUs in highest and lowest quality representations to limit the search process, encoding the highest and lowest quality segments first, and using their block structures and reference frames to shorten the Rate Distortion Optimization process for intermediate quality segments, thereby reducing encoding time-complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If HEVC encoding is used to generate multiple representations at different bit-rates and resolutions, then video quality and compression efficiency are improved, but encoding time-complexity increases significantly
Solution Approach 1:
The patent performs preliminary encoding of the highest quality representation first, then uses its block structure and reference frames as templates for encoding lower quality representations. This preliminary action eliminates the need to perform full Rate Distortion Optimization searches for intermediate quality levels, significantly reducing encoding time while maintaining quality.
Solution Approach 2:
The patent copies the block structure (CTU depth, CU size, PU mode) and reference frame selections from the highest quality encoding to the lower quality encodings. By copying these structural elements and only performing minimal adjustments, the patent avoids redundant computational searches while preserving encoding effectiveness.
2Manufacturing precision
If multiple representations are encoded independently with optimal CU depth structure and PU mode decisions, then encoding precision is improved, but device complexity increases
Solution Approach 1:
The patent merges the encoding process by performing the computationally intensive Rate Distortion Optimization and block structure determination only once for the highest quality representation. The same encoder instance and configuration are then reused for lower quality representations, sharing computational resources and reducing overall device complexity.
Solution Approach 2:
The highest quality encoding serves multiple functions: it provides the reference block structure for all lower quality encodings, establishes reference frames for motion estimation, and defines the overall encoding parameters. This universal reference encoding eliminates the need for separate optimization processes for each quality level.
3Manufacturing precision
If the search process for optimal block structure is performed for all quality levels, then encoding precision is improved, but productivity decreases
Solution Approach 1:
The patent skips the time-consuming Rate Distortion Optimization search process for intermediate quality levels by directly copying the block structure from the highest quality encoding. This skipping of redundant searches dramatically increases encoding throughput while the copied structure maintains sufficient optimization for practical purposes.
Solution Approach 2:
The block structure optimization is performed preliminarily for the highest quality representation, and this pre-optimized structure is then reused for all lower quality encodings. This preliminary optimization eliminates the need for repeated searches, significantly boosting encoding productivity.
Data Source
AI summary
According to embodiments of the disclosure, information of higher and lower quality encoded video segments is used to limit Rate-Distortion Optimization (RDO) for each Coding Unit Tree (CTU). A method first encodes the highest bit-rate segment and consequently uses it to encode the lowest bit-rate video segment. Block structure and selected reference frame of both highest and lowest bit-rate video segments are used to predict and shorten RDO process for each CTU in middle bit-rates. The method delays just one frame using parallel processing. This approach provides time-complexity reduction compared to the reference software for middle bit-rates while degradation is negligible.