An Improved Facial Recognition Technique Using Scale and Rotation Invariant Statistical Moments
Yaser Daanial Khan · 2019
Forensic experts face substantial encumbrance in finding a criminal in electronic databases given a pencil sketch. The problem complexity greatly surmounts when the image database contains thousands or even millions of images. The general approach is to use the sketch images to find the nearest possible match present in the pre-populated database of images consisting of criminals or individuals and groups in their daily routines. The proposed system present an optimal approach to automate the review and search process based on various image analysis techniques. The mathematical computations on images to find the features help us in identifying the characteristics and features of the face. The dataset stored is pre-populated by feature vector corresponding to each image in image database for fast retrieval. Through the experimental run on all the competitive techniques along with the proposed one, it is concluded that the overall accuracy of the systems is 96.2%.