Fake Image Detection by Illumination Color Classification

As images can fetch more information than words, most of the people are dependent on them for communication and use them as powerful evidence in courts and media companies. Nowadays, however, advanced image processing tools and graphical editing techniques makes it easier to edit and modify the digital image. This diminishes our trust in photographs and objects pictures as evidence for real world concept. To maintain the authenticity that the image is not a mixed or spliced image, this paper proposes a new method of detecting digital image forgeries using illumination color classification. Image splicing is a common type of image manipulation operation. To maintain the consistency of the image as well as identifying the areas of manipulation on images without the need of an external support or human process image contents is now days becoming the challenging research problem. In this paper, we are concentrating on authenticity of images that are based on concept of using illumination color estimation. To achieve this, illuminant color is calculated using the physics based method as well as statistical based edge method on similar image region. The estimate of illuminant color is extracted individually from the different face and non-face regions of the image. For the classification performance Support Vector Machine (SVM) approach is used. This technique yields maximum detection rates of 86%. KeywordDigital Image, Illuminant Color Classification, Image Splicing, Forgery Detection, Support Vector Machine (SVM)