We combine pruned tree-structured vector quantization (pruned TSVQ) wi
th Itoh's universal noiseless coder. By combining pruned TSVQ with uni
versal noiseless coding, we benefit from the ''successive approximatio
n'' capabilities of TSVQ, thereby allowing progressive transmission of
images, while retaining the ability to noiselessly encode images of u
nknown statistics in a provably asymptotically optimal fashion. Noisel
ess compression results are comparable to Ziv-Lempel and arithmetic co
ding for both images and finely quantized Gaussian sources.