Improving Performance of Source-Camera Identification By Suppressing Peaks and Eliminating Low-Frequency Defects of Reference SPN

The sensor pattern noise has been broadly used to uniquely recognize digital imaging gadgets. However, the presence of some peaks in Fourier domain and low-frequency defects that are shared among cameras due to the same or comparable in-camera processing strategies leads to increasing the false acceptance rate. In this way, it is important to eliminate these undesirable artifacts to enhance the exactness and reliability. In this letter, we have developed a method for preprocessing the camera reference pattern noise (RPN). This letter is motivated by the fact that refraction of light on dust particle and optical surface affect the major part of camera-RPN. These components are combined and termed as “doughnut” patterns. They are of low frequency in nature. As these low-frequency defects are not characteristic of the sensor, hence they should be removed from the estimated camera-RPN. To achieve this, the proposed method first uses the widely accepted spectrum-equalization algorithm (SEA) to find and suppress the peaks present in the camera-RPN and then eliminates the low-frequency defects in the discrete cosine transform domain. Experimental results performed on the freely available “Dresden” image database show that the proposed method is able to improve the efficiency of the SEA methods, as well as this combination is able to work far better than the other camera-RPN enhancement techniques.