Automated segmentation of skin-tone regions in video sequences

The paper introduces an automatic video object extraction algorithm based on color segmentation. The algorithm focuses on extracting the skin of a human object in an image. The algorithm includes preprocessing, edge detection, classification of pixels and blocks, block clustering, region growing, edge following and final extraction of the human skin-tone region. Through use of these steps, pixel-wise accuracy is achieved in the segmented object of interest. The algorithm gives very promising result in situations such as video conferencing, seminars and surveillance systems.