Face Detection Using Chrominance Boundary Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to accurately detect face candidate areas when the background contains colors similar to skin tones or when the light color temperature is low, leading to erroneous detection of non-face areas as face candidates.
Innovation Solution
The method generates chrominance image information, calculates average chrominance values for target and peripheral areas, evaluates color information to match predefined skin color characteristics, and assesses the difference between these areas to determine if the target area contains a face, using a weighted decision based on the areal ratio and color difference distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If face determination processing is executed on skin color areas to detect face candidate areas, then face detection capability is improved, but erroneous detection of non-face areas as face candidates increases when background colors resemble skin tones
Solution Approach 1:
The patent applies local quality by examining the chrominance characteristics of the determination target area in relation to its peripheral areas. Instead of uniformly evaluating all skin-colored areas, the method locally compares the chrominance values between the target area and its surroundings to identify areas with distinct chrominance boundaries, thereby distinguishing actual faces from background regions with similar colors.
Solution Approach 2:
The patent introduces a new dimension of evaluation by comparing chrominance values between the determination target area and peripheral areas. This spatial relationship dimension complements the traditional skin color detection, creating a two-dimensional evaluation framework (color matching + spatial chrominance difference) that reduces false positives while maintaining detection accuracy.
2Reliability
If chrominance comparison with peripheral areas is performed to reduce false detections, then detection reliability is improved, but processing complexity increases
Solution Approach 1:
The patent segments the image processing into distinct functional modules: chrominance image generation, determination target area setting, average chrominance value calculation for target and peripheral areas, color information evaluation, and final determination. This segmentation allows each module to perform a specific task efficiently, reducing overall processing complexity while maintaining high reliability.
Solution Approach 2:
The patent performs chrominance comparison only for areas that have already been identified as skin-colored candidates. This partial action approach avoids unnecessary computations on non-candidate areas, reducing processing complexity while still achieving high reliability through targeted chrominance boundary evaluation.
Data Source
AI summary
Chrominance image information is generated based upon input image information. A determination target area to undergo determination of a target object within a determination target image constituted with the chrominance image information is set. An average chrominance value for each of the determination target area and a plurality of peripheral areas set around the determination target area is calculated based upon the chrominance image information corresponding to each area. Color information expressed by the average chrominance value calculated for the determination target area is evaluated to determine whether or not the color information matches a characteristic color defined in advance in correspondence to the target object. A difference between the average chrominance value in the determination target area and the average chrominance values in the peripheral areas is evaluated to determine whether or not the determination target area is an image area separate from the peripheral areas. A decision is made as to whether or not the determination target area contains the target object based upon results of an evaluation indicating whether or not the average chrominance value in the determination target area matches the characteristic color and the result of an evaluation indicating whether or not the determination target area is an image area separate from the peripheral areas.