Region Segmentation for Facial Image Compression

This paper addresses the segmentation of passport images in order to improve quality of significant regions and to further reduce redundancy of insignificant ones. The approach is to first segment a facial image into two major regions, namely background and foreground. Here a new technique using pixel difference is presented. To compress facial regions at better quality, a face segmentation algorithm is introduced that detects eyes and mouth in a face. Region of interest (ROI) coding is then used to obtain better quality for facial features. At the end, some strategies that make use of region segmentation are proposed in order to increase performance in entropy coding