Motion Detection Using Binary and Multivalued Image Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion detection methods in video encoding, such as H.26x, MPEG, and HEVC, face high calculation costs and reduced search accuracy due to full search methods, particularly in flat image portions when using binary or N-valued images, which affects encoding efficiency.
Innovation Solution
A motion detecting apparatus that generates images with fewer gradations than the original, performs motion detection using these images, calculates feature parameters, and selects between detected motion vectors and reference vectors to improve accuracy while reducing calculation costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full search is performed with multivalued image to obtain accurate motion vector, then search accuracy is improved, but calculation cost increases significantly
Solution Approach 1:
The patent divides the motion search process into two stages: first performs motion search on binary images to obtain preliminary motion vectors, then uses these vectors to guide a second motion search on multivalued images within a reduced search area. This segmentation allows the computationally intensive multivalued image search to be performed only in limited regions, significantly reducing total calculation cost while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary motion search using binary images before conducting the final motion search with multivalued images. The binary image search provides initial motion vector estimates that serve as starting points for the subsequent multivalued image search, eliminating the need to search the entire search area and thereby reducing calculation cost.
2Power
If binary image is used for motion search to reduce calculation cost, then calculation cost is reduced, but search accuracy deteriorates in flat image portions
Solution Approach 1:
The patent uses binary images as an intermediary tool to obtain preliminary motion vectors, which then guide the search in multivalued images. The binary image search acts as a mediator that provides initial estimates, allowing the system to avoid direct full-search on multivalued images while still achieving accurate motion vectors through the subsequent refined search.
Solution Approach 2:
The patent applies different image types to different stages of the search process: binary images are used for the initial broad search to reduce calculation cost, while multivalued images are used for the final localized search to ensure accuracy. This local quality approach optimizes the use of computational resources at each stage.
3Power
If search area is reduced for multivalued image search to lower calculation cost, then calculation cost is reduced, but search accuracy deteriorates when motion vector is significantly deviated
Solution Approach 1:
The patent performs preliminary motion search using binary images to obtain initial motion vector estimates before conducting the multivalued image search. These preliminary vectors serve as accurate starting points that define the reduced search area, ensuring that even when motion is significant, the search begins from correct positions and only needs to explore local variations.
Data Source
AI summary
In a motion detecting apparatus which can perform a high-accuracy motion search with a less calculation cost, it is discriminated whether or not input Activity and DIFF_XORSUM values are smaller than respective predetermined thresholds. If the Activity value is smaller than the threshold or the DIFF_XORSUM value is smaller than the threshold, a reference vector is selected as a motion vector. On the other hand, if the Activity value is equal to or higher than the threshold and the DIFF_XORSUM value is equal to or higher than the threshold, an N-valued image search vector is selected as the motion vector.