Motion Estimation via Inhomogeneity Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion estimation techniques for image sequences are inefficient and costly, making them unsuitable for commercial portable devices, and the stabilization methods based on these techniques are inadequate due to high computational complexity and poor performance.
Innovation Solution
A method for estimating a global motion vector is developed by subdividing image regions into subregions, assigning weighting coefficients based on inhomogeneity measures, and using these coefficients to calculate the motion vector, which discards unreliable blocks and weights reliable ones to obtain a representative image motion vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional motion estimation techniques are used, then motion compensation can be achieved, but computational complexity and processing cost become excessively high
Solution Approach 1:
The patent divides the image into multiple blocks and further subdivides each block into smaller subregions. This segmentation allows the motion estimation to be performed locally on smaller units rather than on the entire image, significantly reducing the computational complexity while maintaining estimation accuracy through the cumulative effect of local motion vectors.
Solution Approach 2:
The patent assigns different weighting coefficients to different subregions based on their inhomogeneity measures. This local quality approach allows the system to focus computational resources on regions that provide more reliable motion information (higher weighting) while reducing or skipping processing in regions with low reliability (lower or zero weighting), thus optimizing the balance between accuracy and computational cost.
2Measurement precision
If all image blocks are processed equally, then comprehensive motion information is captured, but computational effort increases unnecessarily
Solution Approach 1:
The patent introduces inhomogeneity measures as a parameter to evaluate and compare different image blocks. By calculating this parameter for each block and using it to determine weighting coefficients, the system dynamically adjusts the processing intensity based on block characteristics, thereby improving processing efficiency without sacrificing the reliability of motion vectors from important regions.
Solution Approach 2:
The patent applies local quality by assigning different weights to different blocks based on their inhomogeneity measures. Blocks with higher inhomogeneity (more reliable motion information) receive higher weights and are processed more thoroughly, while blocks with lower inhomogeneity receive lower weights or are skipped, optimizing the balance between measurement precision and productivity.
3Quantity of substance
If unreliable blocks are included in motion estimation, then more data is utilized, but estimation accuracy deteriorates
Solution Approach 1:
The patent discards or down-weights unreliable blocks (those with low inhomogeneity measures) from the motion estimation process. By selectively excluding these blocks rather than including them with equal weight, the system maintains estimation accuracy while still utilizing data from reliable regions. The weighting mechanism recovers the contribution of useful blocks while filtering out harmful noise from unreliable blocks.
Solution Approach 2:
The patent uses inhomogeneity measures as a filtering parameter to distinguish between reliable and unreliable blocks. By changing the inclusion criterion from a simple count of blocks to a weighted sum based on inhomogeneity measures, the system ensures that only blocks with sufficient reliability contribute to the motion estimation, maintaining accuracy while utilizing appropriate data quantity.
Data Source
AI summary
A method of estimating a global motion vector representative of the motion of a first digital image with respect to a second digital image, the first and the second image forming part of a sequence of images and being made up of, respectively, a first and a second pixel matrix. The method estimates the global motion vector on the basis of the estimate of at least one motion vector of at least one region of the first image representative of the motion of the at least one region from the first image to the second image and comprising phases of: subdividing the at least one region of the first image into a plurality of pixel blocks, assigning to each block of the plurality a respective weighting coefficient calculated on the basis of a respective inhomogeneity measure, and estimating the at least one motion vector of said at least one region on the basis of the weighting coefficients assigned to each block of the at least one region.