Description

The source camera identification is enhanced by Auto-White Balance mechanism. White balance is defined as the problem of disentangling the effects of the light source from the resulting RGB-image without changing the actual contents of the image. AWB provides re-balanced image to compute Image quality metrics feature vectors. To determine the Source camera model, Support Vector Machine (SVM) uses RBF kernel. The problem is to identify the cameras of same model by making use of AWB reconstructed images. Based on the ‘idempotence property’ (which produces the same output if executed once or multiple times) of white balance method ,various AWB approaches on illuminant estimation such as Gray World, White-Patch, max-RGB, Shades of Gray, Gray Edge algorithms are implemented. Auto white balance is chosen because it is not been used in image forensics. The above mentioned algorithms are then subjected to image quality metrics feature extraction. The extracted features are of higher dimension. To reduce it to lower dimension, Sequential Backward Selection algorithm is implemented. This method attempts to optimize some criterion by removing features from initial candidate feature set. The dimension reduced features are fed to Support Vector Machines for classification. The classifier makes use of RBF kernel with the true labels to predict the test image features of camera models.