The invention discloses a self-adaptive compressed sensing reconstruction method based on image information content difference. The method comprises the following steps: firstly, carrying out block processing to an inputted original image; then, calculating the sampling factor of the image blocks according to the defined image information content difference coefficient; according to the value of the sampling factor, using the proposed self-adaptive image block sampling and allocating scheme, calculating the number of samplings corresponding to the image blocks; when the image is sampled at the acquisition end, using a part of the Hadamard matrix as the measurement matrix of the image blocks; and conducting sampling according to the number of samplings allocated to each image block; and finally, using the orthogonal matching and tracking algorithm and other reconstruction algorithm based on the sampled image signal at the decoding end, reconstructing the image blocks and combining them in a certain mode for a complete image. Compared with the existing image processing method using the same sampling rate for the image blocks, the method of the invention can effectively use the image information content difference to sample the image blocks, effectively optimize the sampling resources and improve the image reconstruction quality.