Skin Detection using Improved ID3 Algorithm

Abstract

Skin detection is classification the pixels of the image into two types of pixels skin and non-skin. Whereas, skin color affected by many issues like various races of people, various ages of people gender type. Some previous researchers attempted to solve these issues by applying a threshold that depends on certain ranges of skin colors. Despite, it is fast and simple implementation, it does not give a high detection for distinguishing all colors of the skin of people. In this paper suggests improved ID3 (Iterative Dichotomiser) to enhance the performance of skin detection. Three color spaces have been used a dataset of RGB obtained from machine learning repository, the University of California Irvine (UCI), RGB color space, HSV color space, and YCbCr color space. The experimental results demonstrate that the proposed system achieves accuracy up to 99.88%, 99.88%, and 99.80% in a dataset of RGB, a dataset of HSV, and a dataset of YCbCr respectively.