In this paper, we propose derivatives of adapted nonlinear multiresolution decompositions, that allow progressive lossless image compression and that seem to be well suited to satellite images. We compare their performance (bit rate, coding time and decoding time) with different lossless codecs for MRI medical images and for satellite images. The multiresolution decompositions presented are based on a hierarchical pyramidal scheme. For one level of decomposition, the image is partitioned into two polyphase components I1 and I2. One component, I2, is linearly estimated from the observation of I1 and a causal part of I2. The filter coefficients are computed in order to minimize either a global or a local mean square error. The difference between I2 and its truncated linear estimation is then partitioned into two polyphase components, by decimation. Finally, the same process (decimations and estimation) is applied to I1 in order to give two other subimages.