Sparse Optical Flow Motion Compensation for Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques face challenges in achieving high compression ratios without sacrificing picture quality, especially in lossy compression, which often results in visible spatial artifacts, particularly at low bit rates.
Innovation Solution
The method estimates a motion vector at a target position by weighting contributing motion vectors obtained from different transformations within a sparse motion field representation, allowing for a more complex motion model to be described with fewer parameters, and predicting optical flow to reduce signalling overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If lossy compression is used to achieve high compression ratios, then compression efficiency is improved, but picture quality deteriorates with visible spatial artifacts
Solution Approach 1:
The patent segments the motion field into multiple regions, each characterized by different motion models (e.g., affine transformations). This allows different parts of the image to be compressed using appropriate models, achieving high compression ratios while preserving picture quality by avoiding uniform compression of heterogeneous motion patterns.
Solution Approach 2:
The patent changes the parameter representation by using affine transformation parameters (6 parameters per region) instead of dense pixel-level motion vectors. This parameter transformation enables efficient compression while maintaining motion information accuracy, resolving the contradiction between compression efficiency and picture quality.
2Manufacturing precision
If a dense motion field representation is used to maintain picture quality, then motion modeling accuracy is improved, but signalling overhead increases
Solution Approach 1:
The patent divides the image into multiple motion regions, each described by a small number of affine transformation parameters. This segmentation approach reduces the total number of parameters needed compared to dense motion field representation, thereby reducing signalling overhead while maintaining accurate motion modeling through region-specific transformations.
Solution Approach 2:
The patent applies different motion models to different regions based on their specific motion characteristics. Each region uses the most appropriate affine transformation model, optimizing the balance between motion modeling accuracy and data efficiency. This local adaptation reduces overall signalling overhead compared to a uniform dense representation.
3Adaptability or versatility
If a complex motion model is used to describe motion within larger areas, then motion modeling capability is improved, but the number of parameters increases
Solution Approach 1:
The patent transforms motion representation into affine transformation parameters, which provide a compact and efficient way to describe complex motion patterns. This parameter transformation enables the system to model diverse motion types (translation, rotation, scaling, shear) using only 6 parameters per region, improving motion modeling capability without proportionally increasing the number of parameters.
Data Source
AI summary
Methods and apparatuses are provided for estimating motion vectors of a dense motion field based on subsampled sparse motion field. The sparse motion field includes two or more motion vectors with their respective start positions. For each of the motion vectors, a transformation is derived which transforms the motion vector from its start point into a target point. The transformed motion vectors then contribute to the estimated motion vector on the target position. The contribution of each motion vector is weighted. Such motion estimation may be readily used for video encoding and decoding.