USING INTEGRATED COLOR AND TEXTURE FEATURES FOR AUTOMATIC HAIR DETECTION

Hair is one of the most challenging facial features, playing important role in human appearance. This paper introduces novel approach to human hair extraction, based on integration of texture, shape and color features. This approach allows robust hair detection in complex background under various illumination and hairstyles. This study deals with color images and hair detection aims hair re-colorization to simulate different hair colors. The process starting with typical facial features extraction such as open skin, eyes and mouth and finishing with fuzzy hair mask. Fuzzy hair representation allows overcoming hair appearance problems around hair roots and close to outer line of hairstyle. Fuzzy hair mask building involves binary hair mask detection, background detection and matting procedure. Experiments caring in cosmetic store environment for 354 images show 75% of correct detection for complex background and illumination and 85% for homogeneous background and illumination.