An improved DCT block based technique for copy-move forgery detection in medical images

Abstract

Copy-Moved forgery is a common method to manipulate images. Several attempts of image forgery have been discovered and involves a region been duplicated and copied and pasted on another region of the same image in other to achieve selfish gain. Generally, there are two classification of copy-move forgery detection technique such as the block-based and key point-based. The most commonly used technique is the block based which divides image into blocks during the stage of image pre-processing before features are extracted whereas key point based technique skips the division of image into blocks and directly extracts different local feature from the image. In this paper, we review various block based and key point approach which has been proposed by various researchers. The proposed technique is based on DCT and an improvement on DCT technique is achieved in terms of dimensionality reduction using an octagonal block to reduce the number of features for matching, thereby improving detection accuracy. Based on the analysis of this work as compared to previous proposed works, since previous work represents about 79% of the quantized DCT coefficients on each image block and this proposed work represents about 85% of quantized DCT coefficients, therefore, recovery of about 6% more features using the octagonal block was observed as the improvement over the previously proposed dimensionality reduction using the circle block.