Abstract: The traditional image steganalysis methods are based on the features extracted from the whole image, while ignoring the differences of the image content. A new JPEG steganalysis algorithm using quad-treebased segmentation is proposed. First, the given images are segmented to sub-images according to the texture complexity. Then, then steganalysis features of each sort of sub-images with the same or close texture complexity are extracted separately to build a classifier. Finally, the steganalysis results of the whole image are obtained by weighted fusing of all the sub-image categories. Experimental results demonstrate that the proposed algorithm exhibits excellent performance and significantly improves the detection accuracy.