Real-Time Motion Detection via Multi-Resolution Vector Extension
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting motion vectors in high-resolution megapixel images require excessive computation time, making real-time processing difficult, especially in systems processing 60 frames or more per second, and necessitate dedicated hardware.
Innovation Solution
A system and method that generate multi-resolution images in parallel, extract characteristic dots, and extend low-resolution motion vectors to high-resolution motion vectors, reducing computation time by using a parallel image generator, characteristic dot extractor, and motion vector detector, with DMA controllers for efficient memory management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If block matching method is used for all blocks of high-resolution megapixel images, then motion vector extraction accuracy is improved, but computation time increases excessively making real-time processing difficult
Solution Approach 1:
The patent divides the high-resolution image into multiple resolution levels (e.g., 1/4 resolution, 1/2 resolution, full resolution) and processes them separately. The block matching method is applied first to low-resolution images where computation is lighter, then results are extended to high-resolution images, thereby segmenting the computational workload to achieve real-time processing while maintaining accuracy.
Solution Approach 2:
The patent introduces a resolution dimension as an additional layer of processing. Instead of processing only at full resolution, the system creates a multi-resolution pyramid and performs block matching across different resolution levels, using lower-resolution results to guide high-resolution motion vector extraction, thus reducing computation time while maintaining precision.
2Measurement precision
If search area is increased to match high-resolution image characteristics, then motion vector detection accuracy is improved, but computation time increases making real-time processing difficult
Solution Approach 1:
The patent segments the search area by processing images at different resolution levels. The search area for high-resolution images is reduced by first performing block matching at lower resolution levels where the search area is smaller, then extending the results to full resolution, thereby reducing overall computation time while maintaining detection accuracy.
Solution Approach 2:
The patent performs preliminary block matching at low-resolution images before proceeding to high-resolution images. This preliminary action establishes motion vectors at a coarser scale that guide subsequent high-resolution processing, reducing the effective search area needed at full resolution and enabling real-time processing.
3Productivity
If dedicated hardware is used to process high-resolution megapixel images, then real-time processing capability is improved, but device complexity and hardware requirements increase
Solution Approach 1:
The patent segments the processing task into multiple resolution levels that can be handled by standard processors. By dividing the computational workload across different resolution tiers, the system achieves real-time processing capability without requiring dedicated hardware, using instead a multi-resolution block matching algorithm that runs on general-purpose CPUs or DSPs.
Solution Approach 2:
The patent creates copies of the image at different resolution levels (1/4, 1/2, full resolution) and processes these copies through the block matching algorithm. This copying approach allows the system to use standard hardware to process multiple versions of the image simultaneously, achieving real-time performance without dedicated hardware while maintaining the accuracy of high-resolution processing.
Data Source
AI summary
A method and a system for detecting a motion in real time. The system includes: a parallel image generator which generates, in parallel, multi-resolution images, the generated multi-resolution images having at least one or more resolution levels; a characteristic dot extractor which extracts a characteristic dot from at least one of the generated multi-resolution images; and a motion vector detector which detects a motion vector from a low-resolution image based on the extracted characteristic dot and extends the detected motion vector of the low-resolution image to a motion vector of high-resolution.