Research On Image Tampering Recognition Based On Illumination Consistency

With the advancement of technology,digital images have become inseparable from our social connection.From newspapers and magazines to periodicals in science and medicine,they rely heavily on digital images.However,with the emergence of many powerful image processing tools(such as Photoshop),the credibility of digital images has been greatly reduced.Only simple operations on digital images(such as stitching,background rendering,area copying,etc.)can change the content and meaning of the image without notice.In recent years,some people have tampered with real images in news reports,academic research,and court evidence in order to confuse audiovisual and achieve self-harm,which has seriously violated the law.Therefore,the research on digital image recognition technology has important practical significance and value.Copying a person or thing cut from a photo to form a new picture on a different photo is a common way to tamper with the image.Because the lighting environment of the characters in the synthesized image is inconsistent,the lighting effects of the characters in the new image are inconsistent.Therefore,inconsistencies in lighting effects can be used as a basis for detecting tampered images.This article studies the tampering image based on the consistency of illumination.The specific content is as follows:First,it analyzes why the light direction characteristics and light source color characteristics in the light characteristics can be used as the basis for identifying tampered images.Several tampering image detection algorithms based on lighting characteristics are studied,and their advantages and disadvantages in applicability are analyzed.Second,a detection method based on the consistency of light directions is proposed to detect the stitched images.According to the object edge information,this method does not need to use the known illumination reflection model or the assumption conditions,thereby improving the generality of the method of estimating the illumination direction.Third,a detection method based on the color consistency of light sources is proposed.This algorithm uses a unified framework of multiple detection algorithms.By dividing the horizontal and vertical bands of the image and combining the horizontal and vertical detection results of the unified frame,it is detected whether the image has been tampered with.The detection method proposed in this paper can effectively detect tampered images,and the positive detection rate is also improved compared with some other methods.