Affine Intra Prediction for Zoom, Rotation, and Image Block Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding technologies are inefficient when handling high-resolution images with motions such as zoom-in, zoom-out, and rotation due to limited motion compensation models that only consider parallel movements.
Innovation Solution
Implementing an affine directional model for intra prediction using multiple control point modes to derive intra prediction modes for current blocks, allowing for improved encoding/decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing motion compensation models are used that only consider parallel movements, then device complexity is reduced, but encoding efficiency deteriorates when handling video data with zoom-in, zoom-out, and rotation motions
Solution Approach 1:
The patent divides the motion model into multiple control points (at least two control points) within the current block, where each control point has its own motion vector. This segmentation allows different regions of the block to have different motion characteristics, enabling accurate representation of zoom-in, zoom-out, and rotation motions while maintaining manageable complexity through localized motion estimation.
Solution Approach 2:
The patent changes the motion model parameters from simple parallel movement vectors to affine transformation parameters that include scaling and rotation components. By using affine motion vectors that incorporate multiple control points with different motion characteristics, the model can represent complex motions (zoom-in, zoom-out, rotation) while the parameters are derived through systematic calculation from control point motion vectors.
2Productivity
If affine motion model-based motion vector prediction is used to handle zoom-in, zoom-out, and rotation motions, then encoding efficiency is improved, but device complexity increases due to four-parameter or six-parameter models
Solution Approach 1:
The patent applies partial affine transformation by using at least two control points instead of requiring the full six-parameter affine model. This partial application provides sufficient capability to represent zoom-in, zoom-out, and rotation motions through controlled point movements, achieving improved encoding efficiency while avoiding the excessive complexity of complete affine transformation with three control points.
Data Source
AI summary
An image encoding/decoding method and apparatus, a recording medium for storing a bitstream, and a transmission method are provided. The image decoding method comprises determining an affine directional model of a current block, deriving an intra prediction mode of the current block using the affine directional model, and generating a prediction block of the current block by performing intra prediction based on the intra prediction mode.