Abstract: Because the multi-scale block compressed sensing algorithm(MS-BCS-SPL) made each sub-band information to block which had the same sampling rate, while this would cause the unreasonable distribution of resources when these blockes had different edge information. Therefore on the basis of MS-BCS-SPL algorithm, we assigned adaptively the total sampling rate to each block of each layer of sub-band using the different edge information and directional of image blockes, to realize adaptive multi-scale block compressed sensing. Experiments show that, with the different sampling rate especially lower sampling rate, this algorithm not only use less samples and conserve resources; but also recovery more high quality images than the original algorithm.