Color Segmentation Exposure Control for Off-Center Subjects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing exposure control methods in imaging systems often result in suboptimal image quality, particularly when capturing images with people, as they fail to appropriately expose the principal object of interest, such as a human face, especially when positioned off-center according to the Golden Ratio.
Innovation Solution
The method involves classifying pixels in an image based on a predetermined color segmentation, generating a noise-reduced pixel classification map, and using this map to calculate exposure control factors for optimal exposure of the image, ensuring the principal object is appropriately exposed by adjusting parameters like shutter speed and aperture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If frame-averaged exposure control is used, then overall scene brightness is balanced, but principal objects (e.g., human faces) may be overexposed or underexposed
Solution Approach 1:
The patent segments the image into different regions based on color classification, identifying skin-toned regions as principal objects. This segmentation allows separate exposure control for principal objects versus background, resolving the contradiction by enabling different exposure treatments for different image portions.
Solution Approach 2:
The patent applies different exposure control strategies to different regions: skin-toned regions receive exposure adjustment based on skin tone characteristics, while other regions follow frame-averaged exposure control. This local differentiation ensures principal objects are properly exposed while maintaining overall scene brightness balance.
2Illumination intensity
If center-weighted exposure control is used, then center region brightness is optimized, but off-center principal objects are inappropriately exposed
Solution Approach 1:
The patent replaces center-weighted exposure control with color-based region identification. Instead of assuming the center contains the principal object, it identifies skin-toned regions anywhere in the frame as principal objects and applies appropriate exposure control to those specific regions regardless of position.
Solution Approach 2:
The patent changes the basis of exposure control from spatial position (center-weighted) to color characteristics (skin tone detection). This parameter change allows the system to locate and properly expose principal objects based on their color properties rather than their position in the frame.
3Adaptability or versatility
If pixel classification is performed without noise reduction, then exposure control responds to all pixel variations, but noise causes false exposure adjustments
Solution Approach 1:
The patent performs noise reduction on pixel classifications before using them for exposure control calculations. This preliminary processing step removes false signals from noise while preserving genuine scene variations, ensuring that subsequent exposure adjustments are based on accurate pixel classification data.
Data Source
Figure 1a~1b
Figure 2
Figure 3a~3b
AI summary
Embodiments include methods of image processing in which pixels of an image are classified according to color. An exposure control operation is performed according to the pixel classifications. In one example, the pixels are classified according to a predetermined segmentation of a color space, based on predicted sensor responses. In another example, the pixels are classified according to a color selected from among at least two colors.