Spatially partitioned lossless image compression in an embedded framework

We present a new method of lossless image compression which has two additional useful properties: (1) a continuous range of lower resolution images (e.g., lossy images) can be extracted from the representation and (2) any desired region of the image can be individually extracted with up to lossless quality. We achieve these additional properties by spatially partitioning a modified version of the embedded zerotree wavelet compression algorithm. Specifically, the proposed approach uses a new form of successive coefficient refinement which reduces its complexity and improves its rate-distortion performance for lossy decompression. We show that the compression performance of the proposed method is only slightly worse than that of the Said and Pearlman (see IEEE Trans. on Image Proc., vol.5, no.9, p.1303-10, 1996) approach which does not offer regional decompression.