Geometric Reference Picture Warping for Global-Motion Inter Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in maintaining prediction efficiency due to reduced similarity between reference and current pictures, particularly when global motion is involved, leading to degradation in image quality and compression performance.
Innovation Solution
The method involves generating a geometrically modified reference picture by modifying the reference picture and using it for inter prediction, along with encoding and decoding processes that include geometric modification information to optimize prediction block selection based on encoding efficiency and rate-distortion cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If inter prediction is performed using a reference picture with global motion, then compression is achieved, but prediction efficiency degrades due to reduced similarity between reference and current pictures
Solution Approach 1:
The patent applies preliminary action by generating a geometrically modified reference picture before performing inter prediction. The geometric modification (including translation, rotation, scaling, and perspective transformation) is performed in advance on the reference picture to align it with the current picture's geometric characteristics, thereby improving prediction accuracy while maintaining compression efficiency
Solution Approach 2:
The patent changes geometric parameters of the reference picture by applying transformation matrices that modify position, orientation, scale, and perspective. These parameter changes adapt the reference picture to match the current picture's geometric properties, resolving the similarity degradation caused by global motion
2Reliability
If a geometrically modified reference picture is generated and used for prediction, then prediction accuracy improves, but encoding complexity increases due to additional geometric modification information
Solution Approach 1:
The patent applies partial action by selectively applying geometric modification only when necessary (e.g., when global motion is detected or when prediction accuracy improvement is significant). The encoder can choose to use geometric modification for some blocks while using standard inter prediction for others, thereby balancing prediction accuracy improvement with encoding complexity
Solution Approach 2:
The geometric modification parameters (translation, rotation, scaling, perspective) are calculated and stored in advance as geometric modification information. This preliminary preparation allows the decoder to reconstruct the geometrically modified reference picture without requiring complex real-time calculations, reducing overall system complexity
Data Source
AI summary
A method and apparatus use a geometric modified image for video encoding/decoding. The encoding method may include: generating a geometric modified reference picture by geometrically modifying a reference picture; generating a prediction block of a current block within an encoding target picture by performing inter prediction by referencing the reference picture or the geometrically modified reference picture; and encoding inter-prediction information of the current block.