Development of Image Segmentation Methods for the Extraction of Human Faces

Abstract

Image segmentation is the process of dividing the image into sub image to extract the newlineobject of interest. Image segmentation is typically used to locate objects and boundaries newlinein images. The practical applications of image segmentation are content-based image newlineretrieval, machine vision, medical imaging, object detection and recognition tasks like newlineface, fingerprint, iris etc. Face detection is an important step in many computer vision newlinesystems including video surveillance, monitoring of driver state for automotive safety, newlineresponsive user interface, immersive virtual environment and face recognition. Face newlinedetection involves finding location of face, mouth, left and right eye. There are lots of newlineresearch issues involved in face detection, as the human face has a high degree of newlinevariability in appearance, illumination, distance from the image device, occlusion, newlinerotation of head in different places, facial expression, and many more. The problem of newlinedetecting tilted face and side face is very difficult task. Multiple pose face images from newlinedifferent angles are difficult to detect. newlineAn exhaustive study of existing image segmentation methods has been done by us to newlineunderstand existing techniques for facial feature extraction and to devise simple and fast newlinealgorithm for face detection. In this thesis, we propose new techniques for extraction of newlinefacial features that gives reasonably good results. The work is concentrated on detecting newlinesingle frontal face, tilted face, multiple face and side face containing both eye features. newlineThe novel approaches and techniques developed by us for Image segmentation for facial newlinefeature and face extraction are as follows:- newline1. ED-BVT Eye Detection using Boundary Value and Template. This method newlinedetects eye in human face by using boundary value analysis. The method is based newlineon rejecting the area outside the boundary and performing further processing newlineusing template based approach on inside boundary values. newline2. FE-DPXY Face Extraction by finding Dark Pixel group and comparing newlineXY-value.