Efficient content-based copy detection using signs of DCT coefficient

Abstract

In this paper, an efficient content-based copy detection (CBCD) method using the signs of the Discrete Cosine Transform (DCT) coefficient (S-DCT) is proposed. The S-DCTs are extracted from DC images. The proposed method uses a small-size feature, i.e., 48 bits/image, nevertheless, it resulted in neither false-negatives nor false positives. The proposed method can handle common alterations in CBCD application, such as contrast and gamma changing, noise addition, resizing, and a slight shifting. It is confirmed that the efficiency and accuracy of the proposed method are superior to those using the MPEG-7 color layout descriptor, one of the smallest image descriptors.