| Title |
Performance Improvement of SIFT-based Copy-move Forgery Detection Using CSLBP Descriptor |
| Authors |
강태안(Tae An Kang) ; 박준영(Jun Young Park) ; 엄일규(Il Kyu Eom) |
| DOI |
https://doi.org/10.5573/ieie.2020.57.5.65 |
| Keywords |
Copy-move forgery; SIFT; CSLBP; Descriptor; Histogram |
| Abstract |
SIFT(scale invariant feature transform) is a method of extracting features that are invariant in size and rotation, and is applied to various detection and recognition fields. Due to these characteristics, SIFT is widely used as a basic transform for detecting copy-move forgery. However, the SIFT-based copy-move forgery detection method has a disadvantage in that the detection performance is deteriorated when the background region is manipulated, when the forgery region is small, or when the image is compressed. In this paper, we propose the SIFT-based copy-move forgery detection method using the CSLBP(center-symmetric local binary pattern) descriptor to overcome this disadvantage. The CSLBP descriptor has a 16 dimensions and is used in addition to the existing 128 dimensional SIFT descriptor. Experimental results for MICC-F220 and CMH datasets show that the proposed method is more than 95% accuracies. In particular, there is no deterioration in performance with respect to the compressed images. |