Inter Prediction Merge Candidates for Low-Memory Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in efficiently processing next-generation video contents with high spatial resolution and frame rate, leading to increased memory storage and processing demands due to inter prediction based on temporal motion information.
Innovation Solution
A method and apparatus for generating a merging candidate list using spatial and temporal merge candidates, with specific rules for adding collocated blocks and determining search orders based on picture order count and reference picture lists, to optimize inter prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If inter prediction based on temporal motion information is used to process next-generation video contents, then video quality and frame rate are improved, but memory storage and processing demands increase
Solution Approach 1:
The patent extracts only the necessary motion information from reference pictures that are already stored in the buffer, rather than processing all reference picture data. By selectively using motion vectors and motion picture indices from previously decoded pictures, the system reduces memory access requirements while maintaining inter prediction functionality for high frame rate processing
Solution Approach 2:
The patent implements asymmetric handling of reference pictures by distinguishing between different picture types (L0, L1, B0, B1 slices) and applying different motion information extraction strategies. This asymmetric approach optimizes memory usage by treating different reference picture lists differently based on their specific requirements for inter prediction
2Productivity
If inter prediction based on temporal motion information is used to process next-generation video contents, then video quality and frame rate are improved, but processing power requirements increase
Solution Approach 1:
The patent performs preliminary actions by pre-decoding and storing reference pictures in a buffer before they are needed for inter prediction. Motion information from these pre-decoded pictures is cached and readily available when processing current pictures, eliminating the need for re-decoding reference frames and reducing real-time processing power requirements
Solution Approach 2:
The system uses its own previously decoded pictures as reference material for inter prediction, making the decoding process self-sufficient. By leveraging motion information from already-decoded frames stored in the buffer, the system serves its own prediction needs without requiring additional external processing resources
3Measurement precision
If motion information from multiple reference pictures is used for inter prediction, then prediction accuracy is improved, but memory access complexity increases
Solution Approach 1:
The patent segments the motion information extraction process by handling different reference picture lists (L0, L1, B0, B1) separately with specific rules for each. This segmentation allows the system to manage multiple reference pictures systematically, reducing memory access complexity through structured, rule-based selection while maintaining accurate prediction by utilizing the appropriate reference picture for each block
Data Source
AI summary
Disclosed are a method for encoding a video signal and a device therefor. Specifically, a method for decoding an image based on an inter prediction mode comprises: generating a merging candidate list by using a spatial merge candidate and a temporal merge candidate of a current block; obtaining a merge index indicating a certain merge candidate in the merging candidate list; and generating a prediction block of the current block by using motion information of the certain merge candidate, wherein the generating of the merging candidate list may include adding a first merge candidate indicating a collocated block specified by motion information of the spatial merge candidate to the merging candidate list if a reference picture of the spatial merge candidate is same as a predetermined picture.