[발표논문] Source camera identification based on interpolation via lens distortion correction

작성자
caudsil
작성일
2017-10-16 08:24
조회
160
Source camera identification based on interpolation via lens distortion correction
- Australian Journal of Forensic Sciences, 2014

Min Gu Hwang, Hyung Ju Park and Dong Hwan Har

Image reproduction is becoming clearer due to the development of digital technology. Currently, diverse image editing software is being developed in order to manage images more easily. During the image editing process, the programs are able to delete or modify the Exchangeable Image File Format (EXIF) files that contain the original image information. Therefore, images without an origin source are widely spread on the website after editing. This matter could affect image analysis due to the original image becoming distorted. Specifically, images used in a court of law must be accurately sourced for evidence; therefore, a digital image EXIF file without deletion or distortion cannot be considered as appropriate evidence. This research endeavors to trace the identification of a digital camera in order to determine the original digital image and also focuses on a lens distortion correction algorithm that is used during the digital imaging process. Lens distortion correction uses a mapping algorithm and currently also uses an interpolation algorithm in order to prevent an aliasing issue as well as a reconstruction issue. Currently, interpolation shows a similar mapping pattern; therefore, we aimed to discover the interpolation evidence for this pattern. We propose a re-interpolation algorithm in order to detect the interpolation pattern and then adjust the same re-interpolation coefficient in two areas; one with interpolation and the second without interpolation. During a Discrete Fourier Transform (DFT), we confirm a frequency character in each area. Based on this result, we are able to produce a final detection map using the differences between the two areas. In other words, the area that has the interpolation caused by mapping is adjusted with a re-interpolation algorithm, and the second area, which has no interpolation, tends to have a unique frequency character. Moreover, we were able to verify the efficiency of the re-interpolation algorithm when the original image was trimmed.

Keywords: digital forensics; source camera; lens distortion correction; image mapping; interpolation; trimming