Effects of the Database on the Performance of the Human Skin Detection Techniques

Abstract

In many human-related image processing applications, skin color is used as a cue to detect human skin. An Image database is assumed an important key in increasing the separability between the skin and non-skin classes. This paper examines if the image database brings this benefit by evaluating two famous image databases (Compaq and ECU) used in skin detection.  An experiment using Multi-Layer Perceptron Neural Network (MLP), which is a universal classifier, is used to compare the overall performance of the two databases. Surprising results indicate that the image database is a very important element in increasing the classification between the skin and non-skin pixels.