Abstract: To effectively detect copyrighted images suffering from copy attacks is one of the hotspots and difficulties in the image copy detection research. Currently, most of copy detection algorithms can successfully resist the common attacks, such as noise, scale, rotation and so on, but they are quite fragile to cropping and translation attacks. In this paper, an image copy detection algorithm based on SIFT features and ordinal measure of the full DCT coefficients is proposed. The algorithm firstly extracts the same content regions from target image and query image based on matched SIFT features; Secondly, the ordinal measures of the low frequency DCT coefficients in the same content regions are used as the image signatures; Finally the image signatures are used to detect the image copies. The matched SIFT features can locate copy regions accurately, so the proposed algorithm can resist cropping and translation attacks. Meanwhile, the ordinal measure of full DCT coefficients is introduced to resist common attacks. Moreover, the flipping compensation strategy used in this algorithm can resist the horizontal and vertical flipping attacks. The experimental results show that the algorithm has higher precision and recall rate than the existing methods in terms of cropping and translation attacks.