Back close

Recursive Block Based Keypoint Matching for Copy Move Image Forgery Detection

Publication Type : Conference Paper

Publisher : 2020 11th International Conference on Computing, Communication and Networking Technologies, ICCCNT 2020, 2020, 9225658

Source : 2020 11th International Conference on Computing, Communication and Networking Technologies, ICCCNT 2020, 2020, 9225658

Campus : Amritapuri

School : School of Computing, School of Engineering

Center : Computer Vision and Robotics, Research & Projects

Department : Computer Science

Verified : Yes

Year : 2020

Abstract : Images are used as legal proof in areas like forensic investigations. The authenticity of an image intended for investigation may adversely affect the result of such investigations, if the image is a manipulated one. In such fields, the detection of image forgery is very critical and sensitive. Among the prevailing methods, Block based methods divide the images into overlapping regular blocks and finds the match between every block of the whole image. This method is found to be more accurate though computationally expensive. Whereas Keypoint based methods compute the keypoints of an image and find the match between those keypoints. If the image is forged by copy move, there will be highest number of keypoint matches between the corresponding regions. This method is found to be computationally efficient but the accuracy is less. The proposed method utilizes the advantages of both keypoint based and block-based forgery detection methods. We identify meaningful irregular blocks and the similarity of such blocks are measured using the number of matched SIFT keypoints. To identify whether the image is forged or not, an adaptive threshold is employed on the number of keypoint matches and judiciously decide whether to go for block based matching strategy or not for each block. We show that the proposed method achieves better detection rate without compromising the merit on the computational complexity of keypoint based forgery detection.

Admissions Apply Now