Developing Efficient and Robust Techniques for Face Recognition in Forensic Applications

Abstract

Face recognition (FR) is considered as one of the important and essential component in newlinebiometric applications. Biometric face recognition can be used in several applications of newlineForensic Face Recognition (FFR). In forensic scenario facial images available from scenes of newlinecrime represent important information in criminal investigation, as they potentially hold the newlineidentity of the people involved in the crime. In these cases images of a crime scene recorded newlineusing surveillance cameras usually have a low resolution and depicted faces may be partly newlineoccluded. During criminal investigations fingerprint or images in these cameras are used as a newlineclue to identify the victims. Many times such clues are not available and police team take the newlinehelp of eyewitness and forensic artists to draw the sketch of a criminal. Many techniques newlinewere developed in the past years for the recognition of face under well controlled lighting newlineconditions but they fail to perform well when there is a variation in lighting conditions. Face newlinerecognition performance is not yet satisfactory in these circumstances. newlineThis dissertation addresses the problem of FR in forensic applications. In FR variation in newlineillumination is a challenging problem. The differences caused by varying illumination are newlinemore significant than differences seen among individuals. To deal with uneven lighting newlineconditions of digital image pre-processing and illumination normalization methods are used newlinebefore extracting the invariant feature of the face. newlineFR using sketches plays a major role in the field of forensic investigation. In most of the newlinecriminal investigation facial sketches drawn by artist upon the consideration of the newlinedescription given by eyewitness is used. Matching face sketches with photos is a difficult and newlinecomplicated job. Face recognition approaches cannot be implemented on sketches as it is newlinesignificantly different from original photo in terms of shadow, texture and shape. In our work newlinewe consider geometrical features and its relationship on face components