SVD Based Source Cell-Phone Identification

We propose a method for the identification of the source cellphone from a given image. The sensors, digital image formation pipeline and color filter interpolation algorithms used in different brands of cell-phones make the image unique to the camera. Our method is based on the analysis of strict and relative linear dependance in image rows and columns by the use of singular value decomposition (SVD). We define SVD based features that summarize macro and micro statistical properties of the images. These features are then used with support vector machines for the identification of the image sources. Experimental results show that we can identify 10 different sources with 93% accuracy.