Foreground Background Separation Using Subject Profile Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for foreground/background separation in digital images, such as depth from defocus and flash/non-flash techniques, are ineffective in outdoor or well-illuminated scenes and often produce incorrect results due to noise and misclassification of objects at different depths.
Innovation Solution
A method that provides a foreground/background map by comparing regions in a digital image with a subject profile, using techniques like depth from defocus, high-frequency coefficients, and luminance levels, and includes processing steps like smoothing, thresholding, and color segmentation to accurately distinguish foreground from background, even in complex scenes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth from defocus algorithm is used to separate foreground from background, then separation can be achieved in theory, but incorrect results occur due to noise and misclassification of objects at different depths
Solution Approach 1:
The patent combines multiple separation techniques (depth from defocus, flash/non-flash, high frequency coefficients) into a unified system that produces complementary maps. These maps are then integrated through logical operations to achieve more reliable foreground/background separation than any single technique could provide alone.
Solution Approach 2:
The system uses the subject profile as a reference to validate and correct the foreground/background maps. Regions are compared against the subject profile, and discrepancies trigger re-evaluation or correction of the separation results, ensuring higher reliability.
2Adaptability or versatility
If flash/non-flash technique is used for foreground/background separation, then separation is independent of scene illumination, but the technique fails in outdoor or well-illuminated scenes
Solution Approach 1:
The patent creates a universal foreground/background separation system that can operate in multiple lighting conditions. By incorporating multiple techniques with different strengths, the system adapts to various scenarios (indoor/outdoor, flash/non-flash conditions) and selects or combines the appropriate methods for each situation.
3Measurement precision
If multiple images are compared to produce foreground/background map, then separation accuracy improves, but processing time and computational complexity increase
Solution Approach 1:
The patent divides the image processing into distinct stages: generating individual foreground/background maps from different techniques, comparing regions against subject profiles, and finally integrating the maps. This segmentation allows for optimized processing at each stage and enables parallel computation where applicable.
Solution Approach 2:
The system processes images at different resolutions and applies different levels of analysis to different regions. Lower resolution versions are used for initial map generation, while full resolution is applied only where needed for final refinement, reducing overall processing time while maintaining accuracy.
Data Source
AI summary
A method for providing improved foreground/background separation in a digital image of a scene is disclosed. The method comprises providing a first map comprising one or more regions provisionally defined as one of foreground or background within the digital image; and providing a subject profile corresponding to a region of interest of the digital image. The provisionally defined regions are compared with the subject profile to determine if any of the regions intersect with the profile region. The definition of one or more of the regions in the map is changed based on the comparison.