Motion Vector Determination via Optical Flow and Pattern Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for real-time computation of motion vectors in video image frames require expensive hardware or intensive software processing, lacking a lightweight, efficient solution for applications like interactive games.
Innovation Solution
A system that combines optical principles, pattern matching, and stochastic aspects to determine motion vectors in real-time using low-cost hardware like set-top boxes, by normalizing pixel values, computing optic flow, and performing pattern matching between image frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If specialized hardware or intensive software processing is used for real-time motion vector computation, then computation accuracy and real-time performance are improved, but system cost and complexity increase
Solution Approach 1:
The patent divides the image into multiple blocks and processes each block independently to compute motion vectors. This segmentation allows parallel processing of different regions, improving real-time computation speed while keeping each processing unit simple enough for standard processors rather than requiring complex specialized hardware.
Solution Approach 2:
The patent performs pattern matching only within a restricted search area defined by optical flow constraints, rather than searching the entire image. This partial action reduces computational load significantly, enabling real-time processing on standard processors while maintaining adequate motion detection accuracy.
2Productivity
If specialized hardware is used for motion vector computation, then real-time performance is improved, but system cost increases
Solution Approach 1:
The patent replaces complex hardware-based motion detection systems with a software implementation running on standard processors. By using optical flow theory and pattern matching algorithms, the system achieves real-time performance without requiring expensive specialized hardware, thereby reducing system cost while maintaining manufacturing simplicity.
Solution Approach 2:
The patent computes optical flow fields as a preliminary step before performing detailed pattern matching. This preliminary action provides motion constraints that restrict the search area, significantly reducing the computational workload for the subsequent pattern matching stage and enabling real-time processing on cost-effective standard processors.
3Measurement precision
If optical flow computation is performed for all pixels, then motion detection precision is improved, but computational load increases
Solution Approach 1:
The patent extracts motion information from a limited set of reference pixels within each block rather than processing all pixels. By selecting specific reference points and computing optical flow only for these points, the system maintains adequate motion detection precision while dramatically reducing computational energy requirements.
Solution Approach 2:
The patent performs complete optical flow computation only for selected reference pixels rather than all pixels in the image. This partial action provides sufficient motion constraints to guide the pattern matching process, achieving acceptable precision with significantly reduced computational energy compared to full-image optical flow computation.
Data Source
AI summary
A computer system for real-time determination of a motion vector comprising an image processor to apply an image processing filter to a normalized frame element of a first image frame, yielding a filtered frame element having associated filtered pixel values, a pixel selector to select a first reference pixel from the filtered frame element having the highest value of the filtered pixel values, an optic flow module to determine a first optic flow applied at a location of the first reference pixel, a pattern matching module to perform pattern matching between the normalized pixel values of the normalized frame element and normalized frame elements of a second image frame, yielding a plurality of pattern matching scores, and a motion vector determiner to determine a motion vector based on a lowest one of the pattern matching scores. Related apparatus and methods are also described.