Inter Frame Video Forgery Detection using Normalized Multi Scale One Level Subtraction

Abstract

Due to the ease of accessibility to image and video processing software, it has become easier to tamper images and videos without leaving any traces. Such malicious tampering, which cannot be grasped by human eye may lead to undesirable social as well as legal problems. Tampered videos may be used to provide false proof in court or mislead the public about the truth in news reports. This paper aims to propose a novel method called Normalized Multi Scale One Level Subtraction (NMOLS) to detect forgery. In our proposed method video sequence is divided into frames and then pixel grey values of each frame computed, after this all these frames are go through proposed approach. Our system will detect insertion and deletion forgery and it will also detect which type of forgery is done on the video. The proposed method has been assessed utilizing a sizably voluminous dataset. Precision and recall rate of insertion forgery detection is 99% and for deletion forgery detection precision and recall rates are 93% and 96% respectively.