An Image Forensics Algorithm for Blur Detection Based on Properties of Sharp Edge Points
Li Xian Wei, Jun Jie Zhu, Xiao Yuan Yang · Advanced materials research · 2011
This paper proposes a algorithm for detecting manual blur on images, which is usually used to remove obvious traces when tamper images. The algorithm first blurs the test image and blocks the both test image and blurred image. Then extracts and compares the sharp edge points in contourlet domain of the two images, so as to detect the suspicious blurred blocks. Furthermore, differences between manual blur and defocus blur can be indicated by our proposed method, and we can find out whether the image has been manual blurred. We establish a rich set of experimental images, and test results show that the average accurate detection rate is high, and the tampered regions can be always located. Our next work is to improve the robustness of the algorithm.