Skin-tone Macroblock detection for video coding

This paper presents a low complexity color-based skin tone detection classifying skin Macroblocks (MBs) in two categories: Region of Interest (ROI) MBs and No Skin MB as non-ROI MBs (MB; 16×16 block of pixels). The classification is based on empirical thresholds applied to the mean of the color components. With this classification and a modified Rate Control (RC), which smoothly assign various levels of quality, a visual quality increase in human faces is obtained. The challenge is to develop a reliable system that works in real time implementations. Our proposal is a low complexity solution that could run over a single-core DSP as a part of an encoder implementation.