In order to improve the poor robustness of the image region copy-move forgery detection algorithm for the subsequent processing, this paper proposes a new blind detection method based on the mixed moments. Firstly, using Gaussian pyramid transform to extract the low-frequency information from the image and dividing the low frequency part into overlapping blocks; Secondly, the eigenvector of block composed by the exponenti-fourier moments and histogram moments is lexicographically sorted; Thirdly, positioning tampered region precisely and quickly according to the Euclidean distance and space distance. The experimental results show that this method can successfully detect the forged part with translation, rotation, scaling and mixed operation tamper when the image is changed by brightness variation and contrast adjustment.