Auto White Balancing and comparison of Support Vector Machine and neural network classifiers in prediction of source camera

Digital visual media is one of the most commonly used means of communication. But, with the use of low-cost editing tools, tampering and counterfeiting visual contents are increasing enormously. In almost all the Image forensic application areas, the device used for capturing the image is of utmost importance as the origin of the particular image can act as a key evidence to substantiate the legitimacy of the images. This pose a need for the image to be authenticated to establish that the given image is been captured by the device in question. Source camera identification is the process of finding out the camera from which a particular image has been taken.The paper proposes to use Auto White Balancing (AWB) along with parameter extraction to identify the source camera. White balancing is the process of removing the color cast in an image without changing the actual contents of the image. The parameters extracted from the images are fed to the classifiers for training as well as prediction of source camera. We have used Support Vector Machine(SVM) and Neural Network(NN) classifiers for the prediction of source camera. Support Vector Machine prediction has an accuracy 99.672% whereas the prediction accuracy of neural network classifier is lower which is only 92.92%.