Journal
MULTIMEDIA TOOLS AND APPLICATIONS
Volume 78, Issue 15, Pages 20739-20763Publisher
SPRINGER
DOI: 10.1007/s11042-019-7342-9
Keywords
Digital image forensics; Region duplication detection; Copy-move forgery; Harris-Laplace; Hessian-Laplace; Bag of word
Categories
Funding
- National Natural Science Foundation of China [U1736118]
- National Key R&D Program of China [2017YFB0802500]
- Natural Science Foundation of Guangdong [2016A030313350]
- Special Funds for Science and Technology Development of Guangdong [2016KZ010103]
- Key Project of Scientific Research Plan of Guangzhou [201804020068]
- Fundamental Research Funds for the Central Universities [16lgjc83, 17lgjc45]
- Science and Technology Planning Project of Guangdong Province [2017A040405051]
- Alibaba Group through Alibaba Innovative Research Program
Ask authors/readers for more resources
Digital image are easy to be tampered by the photo editing software. Therefore, digital image forensics which aims at validating the authenticity of the digital image are received wide public concern. Region duplication is a commonly used operation in digital image forgeries. The main aims of the region duplication are to overemphasize or conceal some contents by duplicating some regions on the image. Most of the region duplication methods can be categorized into two main classes:block-based and keypoint-based methods. In this paper, a novel region duplication detection scheme is proposed based on hybrid feature and evaluative clustering. The proposed scheme is divided into two stages: the rough matching and the exact matching. In the rough matching, first, hybrid keypoints are extracted from the input image, and those keypoints are described by the unified descriptors. Second, those keypoints are matched by the g2NN strategy. Third, those matched keypoints are grouped by the proposed clustering based on evaluation. Fourth, affine transformations are estimated between these groups, and Bag of Word is used to filter inaccuracy affine transformations to improve the results of pixel level. When no affine transformation is obtained, in the exact matching, each suspicious region is handled separately. Experimental results indicate that the proposed scheme outperforms the state-of-the-art methods under various conditions.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available