Motion Compensated Picture Interpolation Without Motion Estimators
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating interpolated motion-compensated pictures, such as those converting a 50 Hz video signal to 100 Hz, are inadequate when dealing with zoom operations, resulting in visual artifacts like holes and conflicts due to the limitations of motion estimators.
Innovation Solution
A method that interpolates motion-compensated pictures from two consecutive source pictures by defining pixel blocks around each pixel in both source images, computing correlation coefficients, and allocating pixel values based on the arithmetic mean values of these blocks, weighted by the correlation coefficients, without relying on motion estimators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a motion estimator is used to compute motion vectors for each pixel, then motion compensation can be achieved, but visual artifacts such as holes and conflicts appear during zoom operations
Solution Approach 1:
The invention extracts and eliminates the problematic motion estimation step from the traditional motion compensation pipeline. Instead of using motion estimators that create artifacts during zoom operations, the patent directly computes pixel values in the interpolated picture by comparing pixel blocks from two consecutive source pictures, thereby taking out the source of visual artifacts while preserving motion compensation functionality
Solution Approach 2:
The invention changes the fundamental parameter being optimized: instead of optimizing motion vector accuracy through motion estimation, it directly optimizes pixel value accuracy by computing weighted averages of pixel blocks from source pictures. This parameter change from motion-based to intensity-based processing eliminates the artifacts caused by motion estimation during zoom operations
2Reliability
If motion vectors are computed for each pixel, then motion compensation is achieved, but computational complexity increases
Solution Approach 1:
The invention segments the processing into fixed-size pixel blocks instead of processing individual pixels. By defining pixel blocks around reference pixels in source pictures and computing correlations at the block level, the patent reduces the number of computations required while maintaining motion compensation accuracy through the block-based correlation coefficient approach
3Productivity
If traditional motion estimation is used for 50 Hz to 100 Hz conversion, then interpolated pictures are generated, but holes and conflicts appear in zoomed regions
Solution Approach 1:
The invention introduces correlation coefficients as an intermediary mechanism to mediate between source pictures and the interpolated picture. These coefficients act as weights that determine the contribution of pixel blocks from source pictures to the interpolated picture, ensuring complete and accurate pixel value determination in all regions including zoomed areas without holes or conflicts
Data Source
AI summary
The present invention relates to a new method for generating motion compensated picture by interpolating two consecutive source pictures without using motion estimator. For a given pixel to be interpolated, all displacements within a given search range between former and next frame are considered and for each displacement the displaced block difference is calculated. The pixel value is then equal to the sum of each averaged pixel pair within the search range weighted by a correlation coefficient function of the corresponding displaced block difference.