Region-of-interest detection using foreground level and saliency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting regions of interest in images face accuracy issues when a region of interest overlaps a background region or when feature amounts between the region of interest and background are similar, leading to decreased detection precision.
Innovation Solution
A region-of-interest detection apparatus that segments an input image into partial regions, sets a background region, calculates overlap rates and feature amounts, computes a foreground level based on these, and uses saliency calculation to identify regions of interest, thereby enhancing detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If background regions are set in the four corners of the input image and feature amounts are extracted from these background regions, then the detection process is simplified and computation is reduced, but the detection accuracy of regions of interest decreases when they overlap with background regions or have similar feature amounts
Solution Approach 1:
The patent divides the image processing into multiple independent modules: a segmentation unit that segments the input image into multiple partial regions, a background region setting unit that identifies background regions, a foreground level calculation unit that computes foreground levels based on feature amount differences, and a saliency calculation unit that calculates saliency values. This segmentation allows each module to perform its specific function independently, resolving the contradiction by maintaining computational efficiency while improving detection accuracy through multi-stage processing.
Solution Approach 2:
The patent introduces a new dimension of analysis by calculating both foreground levels (based on feature amount differences from background) and saliency values (based on local feature variations) for each position in the image. By combining these two different types of measurements, the system achieves more accurate region of interest detection without significantly increasing overall computational complexity, as each calculation can be performed independently and in parallel.
2Productivity
If feature amounts are compared between background regions and each position in the image to calculate foreground levels, then the detection method is computationally efficient, but detection accuracy decreases when feature amounts of region of interest and background region are similar
Solution Approach 1:
The patent merges two different detection approaches: foreground level calculation (comparing feature amounts between background regions and each position) and saliency calculation (measuring local feature variations). By combining these two complementary methods, the system maintains computational efficiency while significantly improving detection accuracy, especially for regions with similar feature amounts to the background. The detection unit integrates both metrics to make the final determination.
Solution Approach 2:
The patent performs preliminary segmentation of the input image into multiple partial regions before conducting feature amount comparisons and saliency calculations. This preliminary action organizes the image data in advance, allowing subsequent processing steps to operate more efficiently and accurately on pre-processed regions, thereby maintaining high processing speed while improving detection precision.
Data Source
AI summary
A region-of-interest detection apparatus for improving detection accuracy of a region of interest in a case where a region of interest overlaps a background region, or in a case where feature amounts of a region of interest and a background region are similar to each other calculates feature amounts of regions where partial regions and a background region set in an input image overlap each other, and based on the calculated feature amounts and a feature amount of each position in the input image, calculates a foreground level of the position in the input image. Then, the region-of-interest detection apparatus detects a region of interest from the input image based on the calculated foreground level and a saliency of the position.