Comparative Analysis of Deep Learning Approaches for Detecting Copy-Move Forgery

The trusted nature of imagery is more easily manipulated in this tech-driven era through copy-move forgery (CMF), when parts of an image are copied and reinserted into the identical frame to trick users. The diversity of visual properties and the complexity of interfering techniques make it difficult to detect such replicas. Currently, there are two major problems facing the present methodologies being used in CMF detection: very expensive processing and poor detection accuracy of complex visual scenarios. In this vein, our article presents a comparative study on the state-of-the-art learning techniques that have been used to address this problem. The current research seeks to outline the limitations of existing CMF detection algorithms which have great tendencies towards false positive signals as well as low precision in recall metrics. Therefore, this study delves deeper into these issues by examining a holistic resolution for their mitigation.