Skin Segmentation of Facial Images in the Compressed Domain with Applications in Cosmetics

In this paper we present a fast and accurate solution for skin segmentation of facial images in the JPEG compressed domain, in order to classify faces for further cosmetics applications. Within the increasing of image resolution in the past few years, the requirements for large storage space and fast processing (as developments directly in the compressed domain) became essential. The segmentation procedure used here is implemented on the compressed blocks, since this not only reduces the computation time by avoiding the decompression before processing, but also can benefit from direct measures of texture available e.g. in the Discrete Cosine Transform (DCT) coefficients domain and on the de-correlation of colour (YUV colour representation). This almost complete, yet reduced dimensionally, feature space makes easier the training and implementation of rather complex classifiers such as the Bayesian classifier with class probabilities modeled by Gaussian mixtures used here. Results on PUTFACE database have demonstrated the accuracy and robustness of the proposed approach.