Martin Ling wrote: > The fact
there's a quality option on the lcms tool at all suggests that it's decompressing the JPEG, operating on the resulting image and then recompressing. This is the most straightforward way of doing it but will inevitably lead to degradation unless knowledge of the structure of the original compressed data is used.
There's really no alternative with typical JPEGs, since 99% of them will have had the RGB converted to YCC to allow color component sub-sampling. The only smart thing such a program can do is estimate the DCT quantization tables from the incoming file, so that it can use the same quality on compression. Graeme Gill.