Eye location in human face images using fuzzy integral

Locating eyes is an important step in face recognition, face tracking and human-computer communication. In this paper, we present a fast, cheap and robust method for locating eyes using fuzzy method. We use dilation and erosion operations to extract the initial locations of the eyes, to which we apply a circular mask to reduce noise and improve the shape of the eye. The image is binarized by an adaptive threshold value and all the blocks are labeled. Then we eliminate the unsatisfactory blocks of being the eyes. Next, we look for the most similar pairs as eyes. However, similarity is a vague concept, so we use fuzzy integral to handle it. Experiments have demonstrated that the proposed method is fast and reliable.