Information Technology Journal1812-56381812-5646Asian Network for Scientific Information10.3923/itj.2010.152.157HuYuan-Yuan NiuXia-Mu 1201091Perceptual image hashing algorithm based on content-preserving property has a promising application in image retrieval and identification. By statistical analysis of the low frequency coefficients of quantized DCT blocks in JPEG compressed images, a strong correlation between the perceptual similar images is observed. The positions of the most significant low frequency coefficients are found to be invariant during common signal processings and even some geometric distortions. Based on these, a JPEG-compatible image hashing algorithm is proposed. It is verified by our experiments that the proposed algorithm not only has a strong robustness of JPEG compression (QF = 5) and common signal processings, but also is robust to geometric distortions of cropping and resize (more than 5%) and rotation (more than 3 degree), which is absent in other hashing algorithms in JPEG compression domain. Since, JPEG decoder is not needed during generating the hashing, it is more convenient and efficient for index and recognition in large image databases or the internet than those in pixel-wise domain.]]>Monga, V. and B.L. Evans,20061534523465Li, H.F., W.W. Song and S.X. Wang,200648794Manjunath, B.S., C. Shekhar, and R. Chellappa,199629627640Lu, C.S. and H.Y.M. Liao,20035161173Lin, C.Y. and S.F. Chang,200111153168Fridrich, J.,19992536540Fridrich, J. and M. Goljan, 20002000pp: 178183Yu, L.J. and S.H. Sun,20062006pp: 131134Yang, S.H. and C.F. Chen,20052005pp: 110114Lakkundi, R., 20062006