A multi-objective particle swarm optimization based threshold approach for skin color detection

摘要

Skin detection plays an important role in a wide range of image processing applications such as face detection, tracking and recognition. Skin color is often considered to be a useful and discriminating image feature for facial area since it provides computationally effective yet, robust to variation in scale, orientation and partial occlusion. One of the simplest and common approaches is to identify a rlXed decision boundary for different color space. Single or multiple threshold values are characterized for each color space components. Consequently an image pixel value falling within these predefined range(s) is classified as a skin pixel. In this paper, multi-objective particle swarm optimization is employed to determine the optimal threshhold ranges of each components in RGB-CbCrCg color spaces. The performance of the scheme was evaluated employing the ECU face and skin database.