2008.02.15 15:14 "RE: [Tiff] bug? JPEGQUALITY seems not to be read correctly", by Ed Grissom

On 15/02/2008, Grissom, Ed <ed.grissom@intergraph.com> wrote:

There is no place in the header for a "quality" setting. This is not something that is in the JPEG spec or the TIFF spec. What is in both specs is a quality table (q-table). This q-table is generated from the q-setting.

It is difficult to recreate the q-setting that was used to create a q-table.

Ah OK, thanks. But tiff could have a tag in the header which hinted at the Q setting which was used to make the table, couldn't it? It would make the operation of things like tiffcp more predictable.

Certainly it could. And I agree that it would make much more sense to users. We just don't have such a tag defined yet.

In my own JPEG code (non-LibTIFF) have had good success faking a q-setting by looking at one certain value in the table. It has a default value of 100, and by seeing how it is changed in the actual table, I can usually compute a reasonably close version of the q-setting that was used to create the table. Of course, if a custom q-table is used, I am way off the mark, but if it is a scaled version of the default table, it works pretty well.

--
ed grissom
ed.grissom@intergraph.com