In order to improve the performance of coverless steganography algorithm against JPEG compression, this paper proposes a texture synthesis steganography scheme based on super-pixel structure and SVM. Firstly, selects a small texture pattern and partition it into overlap block, in addition, each image candidate block is super-pixel segmented, and the average pixel value of each super-pixel block is used as the feature value of this image candidate block which has structured information. Then use the trained SVM classifier to determine the categories to classify into different collections, each collection maps a secret data. Secondly, when hiding the secret information, the pseudo-random sequence is generated with specified key to determine the position of the texture candidate block placed on the white paper, the candidate block is randomly selected from the corresponding set according to the content of the secret information and is placed on the designated position of the white paper, meanwhile, and the remaining blank areas are filled with texture synthesis method. When extracting information, it is no need to restore the original texture patterns, the secret message is picked according to find specified image block and determine which category it is. Experimental results show that the carrier image generated by this method has good invisibility and better effect to anti-compression.