Video Encoder and Decoder Buffer Control for Parallel Picture Output
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video codecs like H.265/HEVC do not efficiently support advanced parallel processing capabilities for video encoding and decoding, particularly in handling CTU processing, slice organization, and error mitigation techniques such as Region of Interest encoding and Reference Picture Resampling.
Innovation Solution
The invention introduces mechanisms for determining whether to output certain pictures preceding dependent random access pictures, managing coded picture buffer delays, and controlling buffer offset information to enhance parallel processing in video encoding and decoding, including the use of supplemental enhancement information to manage picture output and buffer delays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CTUs are processed in raster scan order with fixed-size blocks, then the encoding process is simple and standardized, but parallel processing capabilities are limited and processing load cannot be effectively balanced
Solution Approach 1:
The patent divides the picture into multiple CTU rows that can be processed in parallel. Each CTU row contains multiple CTUs arranged horizontally, and the entire picture is divided into multiple such rows that can be processed simultaneously by different processing units, thereby enabling parallel processing while maintaining manageable complexity through structured segmentation.
Solution Approach 2:
The patent introduces a new dimension of organization by arranging CTUs into CTU rows and processing them in a raster scan pattern across multiple rows. This dimensional reorganization enables parallel processing across different rows while maintaining the hierarchical structure of CTUs within each row, thus improving productivity without proportionally increasing complexity.
2Reliability
If slices are used for segmentation of coded data, then error mitigation and processing load balancing are improved, but buffer management complexity increases
Solution Approach 1:
The patent uses slices to divide the picture into multiple segments that can be processed independently. Each slice contains one or more CTU rows and can be processed separately, enabling error mitigation where errors in one slice do not propagate to other slices, and allowing processing load balancing across different slices.
Solution Approach 2:
The patent employs preliminary action by pre-defining the slice structure and CTU row organization before processing begins. This allows the decoder to know in advance how the picture is segmented, enabling proper buffer management and reducing complexity during actual decoding by having the structure already established.
3Productivity
If dependent random access pictures are used for advanced parallel processing, then processing efficiency is improved, but control over picture output and buffer timing becomes more complex
Solution Approach 1:
The patent introduces dynamic control mechanisms that allow flexible management of picture output timing. The system can dynamically determine which pictures to output and when, based on the dependent random access structure, enabling efficient parallel processing while adapting buffer timing control to the specific needs of each picture and sequence.
Solution Approach 2:
The patent employs feedback mechanisms where the decoder receives information about the video sequence structure and adjusts its picture output and buffer management accordingly. This feedback loop allows the system to optimize processing efficiency while maintaining proper buffer timing control by continuously adapting to the actual video content and structure.
Data Source
AI summary
An apparatus (200) for receiving an input video data stream according to an embodiment is provided. The input video data stream has a video encoded thereinto. The apparatus (200) is configured to generate an output video data stream from the input video data stream.