Abstract—Human skin color is a useful feature for several computer vision research, including face recognition. This paper presents a human skin image database called SFA. It aims to assist research that use skin color or texture as a feature. SFA was constructed based on face images of FERET (876 images) and AR (242 images) databases, from which skin and non-skin samples and the ground truths of skin detection were retrieved. The samples vary of dimension, from 1 pixel to 35x35 pixels. For each dimension, SFA has 3354 samples of skin and 5590 samples of non-skin. The samples validation was made by a pattern classification process using artificial neural networks, reaching around 93% accuracy. A comparison with UCI database concerning image segmentation was made, where SFA showed almost 4% of improvement. The SFA image database showed great potential to assist research which have skin color or texture as a relevant feature.
Citation Required: CASATI, J. P. B. ; MORAES, D. R. ; RODRIGUES, E. L. L. . SFA: A Human Skin Image Database based on FERET and AR Facial Images. In: IX Workshop de Visão Computacional, 2013, Rio de Janeiro. Anais do VIII Workshop de Visão Computacional, 2013.
QR Code for this URL: 