NALU Header Segmentation for Video Coding Adaptability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The H.264 NAL unit header is insufficient for certain applications, while the H.265 header is unnecessarily long for others, lacking support for modern technologies like 360 video coding, and requiring a design that balances compactness with efficient support for modern applications.
Innovation Solution
The proposed solution introduces a Network Abstraction Layer Unit (NALU) header design that uses a NALU type class and conditional presence of syntax elements to reduce the header length for common NAL units, while allowing additional octets for less common types, incorporating a picture_id syntax element for improved error resilience and bitstream manipulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If H.264 NAL unit header design is used, then the header can be 1 octet in length for certain types, but the header structure is insufficient for modern applications like 360 video coding
Solution Approach 1:
The NAL unit type field is segmented into two parts: a NAL unit type class (4 bits) that identifies the general category, and an extended NAL unit type field that provides additional specificity. This segmentation allows the header to support both traditional H.264 applications and modern applications like 360 video coding without requiring a complete restructuring of the header format.
Solution Approach 2:
The extended NAL unit type field is nested within the existing NAL unit header structure, allowing additional information to be incorporated without disrupting the original H.264 header format. This nesting enables backward compatibility while providing extended functionality for modern applications.
2Ease of manufacture
If H.265 NAL unit header design is used, then support for scalable and Multiview extensions is simplified, but the header length becomes fixed at 2 octets which is unnecessarily long for common NAL units
Solution Approach 1:
The NAL unit header length is made dynamic rather than fixed. For common NAL unit types, the header remains compact at 1 octet, while for extended types requiring modern application support, the header can expand to include the extended NAL unit type field. This dynamic sizing optimizes coding efficiency for common cases while providing necessary support for modern applications when needed.
3Ease of operation
If fixed length codewords are used in NAL unit header, then Media Aware Network Elements can easily manipulate bitstreams, but the header cannot be optimized for different NAL unit types
Solution Approach 1:
The NAL unit header uses local quality by applying different coding strategies to different parts of the header. The NAL unit type class uses fixed length coding for easy manipulation, while the extended NAL unit type field uses variable length coding to optimize for specific NAL unit types. This allows Media Aware Network Elements to easily manipulate the common parts of the header while enabling coding efficiency optimizations for specific types.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A method and apparatus for decoding a video sequence include decoding a fixed length binary coded network abstraction layer unit (NALU) class type included in an NALU header. An NALU type in the NALU header is decoded. A picture is reconstructed, and a type of the picture is identified by a combination of the NALU class type and the NALU type.