Source camera identification based on CFA interpolation

Abstract

In this work, we focus our interest on blind source camera identification problem by extending our results in the direction of [1]. The interpolation in the color surface of an image due to the use of a color filter array (CFA) forms the basis of the paper. We propose to identify the source camera of an image based on traces of the proprietary interpolation algorithm deployed by a digital camera. For this purpose, a set of image characteristics are defined and then used in conjunction with a support vector machine based multi-class classifier to determine the originating digital camera. We also provide initial results on identifying source among two and three digital cameras.

Source Camera Identification Using Support Vector Machines

“…Thus, the interpolation operation can be modeled as a weighted linear combination of neighbor pixels in RGB channels [1,14,15]. For example, a missing G pixel g x,y is interpolated using an n × n neighborhood as:…”

“…Several researchers have employed multi-class classifiers for camera identification [1,4,8,11,14]. The methodology involves extracting feature vectors from several image samples created by various camera models.…”

“…A variety of CFA interpolation patterns are used; the most common is the Bayer pattern. Bayram, et al [1] employ an expectation-maximization algorithm to extract the spectral relationship introduced by interpolation to build a camera-brand classifier. Long and Huang [9] and Swaminathan, et al [14,15] have developed CFA interpolation coefficient estimation methods based on the quadratic pixel correlation model and the minimization problem.…”

New Features to Identify Computer Generated Images

“…The interpolation process can be approximated as a linear filtering and can be characterized in terms of the filter coefficients. In [5,6], it is showed that source digital camera-model can be detected by estimating CFA demosaicking filter coefficients. Since most real images are created by digital cameras, it is expected that they exhibit traces of CFA interpolation, whereas PRCG images should not.…”

A pattern classification framework for theoretical analysis of component forensics

“…Related prior work mostly aim at developing techniques for forensic analysis to estimate the parameters of the different components in the information processing chain. In literature, methods have been proposed to estimate in-camera processing such as color interpolation [1,3,4] and white balancing [5], and post-camera manipulations like resampling [6], lighting, luminance, brightness change [6], and JPEG compression [7]. These collection of prior art provides algorithms to estimate the parameters of many types of in-camera and post-camera processing.…”