Facial image database for law enforcement application: an implementation

Abstract

This paper described an automatic facial feature extraction method from mug shots. Since most facial features locate at specific regions on a facial image, the region detection and partitioning techniques were used to segment and extract facial features. Heuristics were developed to detect the top, bottom, left and right margins of each feature region from the histograms of the vertical and horizontal projections of a facial image. Each facial feature region was then segmented according to its margins. Furthermore, each facial image was transformed to a facial feature vector, of which each element is the angle between two facial feature regions. The Euclidean distance was used to measure the similarities between facial feature vectors. A facial image database consists of three hundred mug shots was used for the experiment. The results show that the proposed scheme is computational efficient and performs well in facial image retrieval.