A robust feature for single image camera identification using local binary patterns

Most of the camera identification methods work in batch mode. However, in many applications, there is just one image from the resource. In addition, previous proposed camera identification algorithms are prepared to classify the sources for a specified compression quality factor. In contrast, the compression qualities of the image may be changed during further processing and tampering. In this paper, a new single image feature is proposed based on local binary pattern of extracted edges from the input image. The feature showed good robustness against JPEG compression quality variations. The feature was empirically explored and showed its discrimination characteristics in Dresden standard camera identification dataset. The method was assessed with state of the art camera identification methods on the dataset and revealed its capabilities in both high and low quality images.