Improving source camera identification performance using DCT based image frequency components dependent sensor pattern noise extraction method

Abstract Sensor imperfections in the form of photo response non-uniformity (PRNU) are widely used to perform various image forensic tasks such as source camera identification, image integrity verification, and device linking. The PRNU contains important information about the sensor in terms of frequency contents, this information makes it suitable for various image forensic applications. The main drawback of existing methods of PRNU extraction is that the extracted PRNU contains fine details of the image i.e., the high-frequency details (edges and texture). For solving this problem we have applied a pre-processing step on widely accepted PRNU extraction methods. Our pre-processing step is based on the fact that ‘PRNU is a very weak noise signal and hence it can be efficiently extracted from the image by applying PRNU extraction method in low frequency (LF) and high frequency (HF) components of the image separately’. Initially, we have applied this pre-processing concept to the widely accepted PRNU extraction methods and found that it is able to improve the performance of most of the PRNU extraction methods. The best improvement takes place for Mihcak filter. Hence in the remaining part of the work, this generalized concept is more precisely applied to the Mihcak filter only. By utilizing the proposed pre-processing idea with the Mihcak filer, the new filter is termed as the pMihcak filter. PRNU extracted using pMihcak filter contains the least amount of HF details of the image. Also, the pMihcak filter is able to extract PRNU from low-frequency components of the image which otherwise not possible for existing PRNU extractors.