Ah, I was thinking of JPEG scanlines. But thanks for the offer of help/info: perhaps you could point me in the right direction? There are times when I need to produce thumbnails from multiple terabytes of data, thousands upon thousands of JPEGs, each up to three meg in size. So I was thinking about producing a faster thumbnailer: in a bitmap image, I'd just build a thumbnail by sampling every n'th byte. I thought I could hack the JPEG library to process only every n'th scanline (easy enough I thought) but I seem to remember the scanlines were compressed. In hindsight this doesn't make much sense for an effective compression algorithm - why restart compression for every line? Have you noticed that (amongst others) the Picassa JPEG renderer initially shows a rough image - I think even for non-interlaced images? This encourages me, perhaps incorrectly, to assume it is somehow doing what I want. Hm.... did that express anything other than my current tiredness? Cheers! lee