Image Content Tampering Detection And Localization Based On Scale Feature Analysis

Due to the progress of computer and image editing software,image forgery is becoming more and more common in our daily life.Some people use advanced image processing software to add or remove parts of the image in a professional way,thus changing its content,resulting in image forgery,which makes it difficult to observe these changes.Therefore,as forgers develop more complex forged images,researchers must design more effective methods to detect these forged images,and the verification of image content has become a very important research topic.Under this background,this dissertation makes an in-depth study on many aspects of image content tamper detection.Considering a fact that a tampered region usually contains scale transformation,an algorithm for image tamper detection and location using scale features is proposed.The main research work of this dissertation is as follows:Firstly,this dissertation introduces the research status of several commonly used tampering operations,and finds that most of the algorithms have poor performance in scale transformation detection.Based on this question,this dissertation expounds the basic principle of scale transformation operation,and then focuses on four typical scale feature algorithms,and compares and analyzes their performance through experiments.Secondly,in view of the fact that SIFT algorithm is a commonly used effective scale feature extraction algorithm,the situation of poor detection performance and inability in locating small tampered areas are studied,and an image copy and paste detection algorithm based on MB-LBP and SIFT feature matching is proposed.The experimental results show that more feature points can be extracted from small tampered areas due to the addition of MB-LBP texture features.And it can also locate the tampered region,but this method only aims at the tampered image with strong texture.Finally,in order to analyze the splicing tampering in the image content tampering operation,the tampery detection is carried out by combining the two scale features at the decision level.This method uses the classical SURF feature and ORB feature to fuse at the decision level,which overcomes the problem of few matching logarithms and rough location when matching based on a single feature.The experimental results show that this method is effective for the detection and location of spliced tampered images.