This paper proposes an adaptive block-based compressed sensing (ABCS) technique to build a new progressive image coding scheme, in which both image acquisition and reconstruction are carried out in two layers. At the base layer, an original image is sampled and restored by the block-based compressed sensing (BCS) method with a low and fixed measurement rate. Second, all blocks in the enhancement layer are re-sampled with different rates according to a block classification. The final reconstruction of a block at the enhancement layer is performed in multiple stages where each stage only knows a part of sampled coefficients. We present some experimental results to show that our proposed ABCS method outperforms the BCS method; in particular, it produces a better visual quality in regions that contain edges, patterns, and textures.