DETECTING DUPLICATE REGIONS IN DIGITAL IMAGES USING IMPROVED LOCALIZATION METHOD

Abstract

Now days, the uses of digital images has increased and so has the forgery mechanisms. Copy-Move forgery is one of the types of image forgery. In the case of copy-move forgery one small region of image is replaced with another region within the same image. Many algorithms already developed for detecting copy-move regions but the major problem for most of the algorithms are localization of duplicate regions within same image after detecting copy-move part. Here we present an efficient approach for detecting duplicate regions in a digital image. In our method, we have used PCA (Principal Component Analysis) to the input image for obtaining compressed image. For obtaining feature vector of each block, we divide the image into number of overlapping blocks. These feature vectors are calculated with the help of principal component analysis. Here we represent eigenvector as feature vector. Lexicographic sorting is henceforth applied to locate the similar feature vector of each block. If two or more feature vectors have same value then this indicates that image forgery has been done. But we cannot guarantee of image forgery because in some cases, feature vectors may be same in an image. Hence, we calculate total number of connected blocks with in the same distance. If the numbers of blocks are greater than a threshold value then image has been forged by duplicate regions, but if numbers of blocks are less than a threshold value then image has not been forged. After that, localization method is used for getting number of pixels that have been copied and pasted. With the help of our proposed method, duplicate regions in an image can be detected more accurately and with less computational complexity as compared to other methods.