Analysis of SIFT Method Based on Swarm Intelligent Algorithms for Copy-Move Forgery Detection

Abstract

Scale Invariant Features Transform (SIFT) method has proven effective for detecting the images with the copy-move forgery in digital forensics field. However, by a great number of tests and practicalities, it is certificated that the detection results highly depend on the presetting of the multiple thresholds. The exhaustive manual searched for the preset thresholds, based on a wise guess, must cause a high computational cost and inefficiency. In this paper, a SIFT method based on swarm intelligent algorithm for copy-move forgery detection is proposed. Three canonical swarm intelligent algorithms (particle swarm optimization-PSO, differential evolution-DE and artificial bee colony-ABC) are applied to find the optimal multiple thresholds for SIFT-based method. Experimental results against various test images with different sizes of duplicated regions show that no algorithm is always more excellent than others. For most cases, PSO algorithm is more adept at finding optimal multiple thresholds for SIFT-based copy-move forgery detection.