Contour Image Generation via Foreground Masking and Edge Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge extraction filters struggle to accurately extract object contours due to sensitivity to shooting conditions and initial parameter settings, and active contour models face challenges in tracking shape changes in moving images due to high computational demands.
Innovation Solution
An image processing apparatus that generates a contour image by extracting edges, using foreground processing to isolate the object region, masking unnecessary edges, and synthesizing the masked edge image with a foreground boundary image to produce a clear contour image, while an object tracking apparatus uses this contour image for likelihood matching and probability distribution estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a general edge extraction filter is used to extract contours, then edge detection can be performed, but minute shadows or designs other than the contour are extracted or the contour is broken and cannot be extracted
Solution Approach 1:
The patent segments the edge extraction process into two distinct parts: (1) general edge extraction using an edge extraction filter to capture all potential edges, and (2) foreground-specific edge extraction using foreground processing to isolate only the edges belonging to the target object. This segmentation allows the system to first capture comprehensive edge information and then filter it down to only the relevant contour edges, resolving the contradiction between comprehensive detection and accurate contour extraction.
Solution Approach 2:
The patent introduces a foreground processing unit as an intermediary between the edge extraction filter and the final contour output. This intermediary component processes the extracted edges through foreground analysis, using techniques such as background subtraction and foreground region identification to distinguish true contour edges from spurious edges caused by shadows or designs. This intermediary step acts as a mediator that refines the raw edge data into accurate contour information.
2Measurement precision
If the threshold for edge extraction is adjusted to improve contour accuracy, then false edges may be reduced, but the edge extraction frequency on the whole screen varies and contours may still not be extracted reliably
Solution Approach 1:
The patent applies local quality by treating different regions of the image differently in the edge extraction process. Instead of using a uniform threshold across the entire image, the system performs foreground processing to identify which regions contain the target object and applies edge extraction selectively to those regions. This allows the system to maintain high extraction accuracy in foreground regions while avoiding false edges in background regions, achieving both local precision and overall consistency.
3Measurement precision
If active contour models are used to extract contours, then contour extraction can be performed, but initial settings for a contour model are required with respect to each object and the final result is affected by the initial settings
Solution Approach 1:
The patent performs preliminary action by conducting foreground processing before the contour extraction step. The foreground processing unit pre-identifies the target object region using background subtraction and foreground analysis, creating a prepared state that guides the subsequent edge extraction. This preliminary identification eliminates the need for complex initial settings in active contour models, as the system already knows which region to focus on before contour extraction begins.
4Reliability
If visual tracking techniques with particle filtering are used to track moving objects, then object tracking can be performed, but the amount of calculation is large and shape changes of the object may not be followed
Solution Approach 1:
The patent extracts only the essential contour edges from the image using foreground processing, rather than processing the entire image or using computationally intensive particle filtering on all image data. By isolating and extracting only the relevant contour information, the system reduces the data volume that needs to be processed for tracking, thereby improving processing speed while maintaining tracking reliability.
Data Source
AI summary
An edge extracting unit 66 of a contour image generator 22 generates an edge image of an input image 62 using an edge extraction filter, etc. A foreground processing unit 68 extracts the foreground from the input image 62 using a background image 64 and expands the foreground to generate an expanded foreground image. The foreground processing unit 68 further generates a foreground boundary image constructed of the boundary of the expanded foreground region. A mask unit 70 masks the edge image using the expanded foreground image to eliminate edges in the background. A synthesis unit 72 synthesizes the masked edge image and the foreground boundary image to generate a contour image.