An Image Forensic Technique Based on 2D Lighting Estimation Using Spherical Harmonic Frames
In this paper, a novel approach for exposing digital image tampering based on the theory of spherical harmonic frames is presented. We describe a robust technique for exposing digital forgeries that we utilize the information along a 2D occluding contour
- PDF / 1,282,024 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 30 Downloads / 188 Views
Abstract. In this paper, a novel approach for exposing digital image tampering based on the theory of spherical harmonic frames is presented. We describe a robust technique for exposing digital forgeries that we utilize the information along a 2D occluding contour and estimate the lighting feature using spherical harmonic frames. Spherical harmonic frames are generated by the rotation along the symmetry axes of a symmetry group. The lighting-based digital forensic technique using spherical harmonic frames inherits the robust property of frames and improve the statistical results compared with spherical harmonic bases. Experimental results performed using spherical harmonic frames prove the robust measurements and discriminability of the complex lighting environments from synthetic data and real data. The application of identifying the tampered images reveals the improvement of our method. Keywords: Spherical harmonic frames Lighting
1
·
Digital forensic technique
·
Introduction
A major effort of the research community has been devoted to the digital forensic technique of exposing image tampering[1–5]. As the recent advances in computational photography, computer vision, and computer graphic, the development of friendly and easy-to-use manipulation tools has made the community into a crisis of confidence, which is caused by pervasive use of digital fakes in legal certification, the media, advertising, entertainment industry, national security, and more. In this new environment, the demand for efficient forensic tools that can accurately and trustfully expose the digital fakes is emerging. This work is supported by Postdoctoral Science Foundation of China No. 2013M540822, National Natural Science Foundation of China(NSFC, No. 61340046, 60875050, 60675025), National High Technology Research and Development Program of China(863 Program, No. 2006AA04Z247), Scientific and Technical Innovation Commission of Shenzhen Municipality (No. JCYJ20120614152234873, CXC201104210010A, JCYJ20130331144631730, JCYJ20130331144716089). c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part I, CCIS 546, pp. 325–334, 2015. DOI: 10.1007/978-3-662-48558-3 33
326
W. Zhao and H. Liu
In the area of image forensic, researchers have successfully developed a variety of image authentication tools according to different situations. When utilizing multiple images to generate a tampered image, since the image targets are captured under different lighting conditions, it is difficult to keep lighting consistency between different targets. Using lighting inconsistency between the different targets of one image to identify image authenticity is one technology of image forensics. In previous work, Johnson and Farid [6] use lighting information extracted from 2D occluding contour to identify the authenticity of the image. In order to extract 3D lighting information in a single image, and they take advantage of the spot position of the human eye to estimate the direction of light [7]. Johnson, Farid [8], Stork [9] and Kee [10], utiliz
Data Loading...
