Scam Image Detection on Copy-Move by JPEG Features and Classical Block Matching with Improved Variant

Abstract

Numerous methods have been developed to identify copy-move forgeries, which are among the most often used alteration strategies of digital photographs. The most widely used format of digital photographs is JPEG, which allows for high-rate compression without drastically altering the meaning of the picture. The objective of this work is to develop a system that can automatically identify the forgery type of the suspect image through in a single procedure, without requiring any kind of expert information. A preferable method is to run the same image through multiple algorithms, which saves time and prevents the needless evaluation of multiple detection results, from which it may be difficult to determine the correct output due to the presence of multiple confounding factors. Additionally, it has been shown that the established method is very effective in detecting expert forgeries when the duplicated region is picked in a non-rigid fashion, which is almost hard for the human eye to perform.