This paper presents a novel multi-resolution image compression scheme based on iterative subband and binary coding. The signal is split into a low frequency image and its high frequency complement. The first is compressed conventionally or by using adaptive sinusoidal models. The high frequency component is treated via a non-linear transformation that leads to a binary image and its complement that can be brought to lower resolution via filtering. The process is iterative and the number of iterations depends on the size of the image and the desired compression rate.