Copy move forgery detection using SIFT and GMM

摘要

Modifying or enhancing an image is ubiquitous but, when enhancement tends to change the interpretation of the image they are termed as an attempt of forgery on digital images. Copy move forgery (CMF) is a simple technique and has a number of well built tools in a number of image enhancement software. CMF detection techniques often tend to establish similarity between copied and pasted region on the same image as both are from same original image. Keypoint and block based techniques are used to determine the CMF. SIFT keypoints are combined with different techniques to accurately localize forgery. High dimensionality of feature vector acts as a bottle neck in SIFT based analysis. We propose a method to detect CMF using SIFT descriptors which are clustered using GMM and segment the obtained suspect region speeding up the analysis.