Image Processing Apparatus for Accurate White Balance Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to accurately estimate the color temperature of a light source from image data, particularly when images are not in color balance, leading to erroneous detection of skin-tone and gray pixels, which can result in inappropriate white balance correction.
Innovation Solution
An image processing apparatus and method that detects skin-tone areas based on image shape or structure, then uses color information from these areas to accurately identify gray pixels, estimating the color temperature of the light source by converting pixel values to chromaticity values and determining their proximity to blackbody loci on a chromaticity diagram.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If color information alone is used to detect skin-tone and gray pixels, then the detection process is simple, but the detection accuracy deteriorates leading to erroneous pixel identification
Solution Approach 1:
The patent segments the detection process into two distinct stages: first detecting skin-tone pixels using color information, then detecting gray pixels using both color information and spatial relationship information. This segmentation allows each stage to use appropriate detection methods, improving overall accuracy while maintaining reasonable complexity
Solution Approach 2:
The patent introduces spatial relationship information as an intermediary factor to mediate between color information and final pixel classification. By incorporating positional relationships and spatial context, the system resolves ambiguities in color-based detection without requiring completely complex alternative methods
2Productivity
If white balance correction is performed without accurate color temperature estimation, then the processing is fast, but the color accuracy deteriorates resulting in inappropriate correction
Solution Approach 1:
The patent performs preliminary detection of skin-tone and gray pixels with high accuracy before the actual white balance correction process. By accurately identifying reference pixels in advance using the dual-criteria method (color + spatial relationship), the subsequent color temperature estimation and correction can proceed efficiently with confidence in the reference data quality
Solution Approach 2:
The patent replaces simple color-threshold-based detection with a more sophisticated detection mechanism that incorporates spatial relationship analysis. This substitution enables more accurate pixel identification that better supports reliable color temperature estimation, improving the foundation for accurate white balance correction
3Loss of time
If images are printed without image processing, then the processing time is minimized, but the color accuracy deteriorates causing tint reflection from photographic light sources
Solution Approach 1:
The patent enables the image processing system to automatically perform accurate white balance correction by self-identifying skin-tone and gray pixels through the proposed detection method. The system uses its own detected reference pixels to calculate color temperature and apply correction, eliminating the need for manual intervention or external calibration while maintaining both speed and accuracy
Data Source
AI summary
A skin-tone image portion contained in an image is detected based upon the shape of the image of a human face. An average value of each of RGB values of pixels that constitute the skin-tone image portion detected is calculated. If the distance between a skin-tone—blackbody locus and a value that is the result of converting the RGB values obtained by multiplying the average value by prescribed coefficients is less than a prescribed value, these coefficients are adopted as coefficients for multiplying the RGB values of each pixel constituting the image. By using a value that is the result of converting, to a chromaticity value, the RGB values obtained by multiplying the RGB values of each of the pixels constituting the image by the prescribed coefficients, those pixels of the image that have values belonging to a zone in the vicinity of a point on a gray—blackbody locus that corresponds to a light-source color temperature estimated based upon the skin-tone—blackbody locus are treated as gray candidate pixels.