Dual-bound-constraints in morphological model refinement for multiple skin color tones detection

摘要

This paper aims to propose a solution to skin color detection problem when existing multiple skin-color target persons in the image. The AdaBoost algorithm is used to detect the position of the face, and the dominant color of every face regions is used as the initial skin color sample. A new type of sample update mechanism is used to increase the skin color detection accuracy rate, with a constraint mechanism controlling the number of iterations, allowing the proposed system to find out a more complete and smooth skin color regions. In comparison to other methods with fixed samples, the skin color samples in this study varies with each execution, and the resistance to the skin color shift caused by light and shade is improved.