Abstract: For many years the usage of photography is done for space-time events, used as evidence in crime scenes. Photography is developed to such a extend that people rely on the phrase that says One picture says thousand words. However every development has pros and cons, hence much powerful editing software for images has been developed thus making the modification and forgery easy and not acted upon. Therefore image composition is increased. Hence forgery of images or splicing is increased which is type of splicing where a fake object or image is added in an original image. After the insertion of the fake object it is hard to determine the original image. Thus the question arises regarding these photographs used in evidence or for any personal usage. To tackle this kind of misuse we propose forgery detection that helps to find fake images based on the illuminated inconsistencies which are subtle in the images of the photographs. This is a machine learning based approach. This technique is available for only the images containing the faces of two or more human only. We incorporate the information from physics and statistical-based illumination. We do estimation in the regions of faces only. Then we extract texture and the edge based features. These estimations are then sent to the machine learning approach and the automatic decision is made.
Keyword: Forensic of image, machine learning, spliced image detection, constancy in color, illuminated color.
Title: Detection of Digital Image Forgeries By Color Classification Based On Illumination
Author: Mansi Bhalerao, Rutuja Bhatambrekar, Mayuri Ghuge, Shradha Katte
International Journal of Computer Science and Information Technology Research
ISSN 2348-120X (online), ISSN 2348-1196 (print)
Research Publish Journals
