Arithmetic coding broken in IJG-code

[thorfdbg]

Just a quick warning: I'm currently doing the JPEG compliance tests using 10918-2 (ITU-T.83) and found that the IJG implementation of the arithmetic coding option defined in 10918-1 is actually non-conforming. The problem here is that the probability contexts need to be additionally conditioned on the component index in the scan, but IJG uses unconditioned contexts. I have fixed this in the upcoming new revision of libjpeg (NIJG - not the IJG code). For the time being, avoid the AC coding option.

[thorfdbg]

Just a quick warning: I'm currently doing the JPEG compliance tests using 10918-2 (ITU-T.83) and found that the IJG implementation of the arithmetic coding option defined in 10918-1 is actually non-conforming. The problem here is that the probability contexts need to be additionally conditioned on the component index in the scan, but IJG uses unconditioned contexts. I have fixed this in the upcoming new revision of libjpeg (NIJG - not the IJG code). For the time being, avoid the AC coding option.

Not quite so - my fault here! The trouble is actually that the context of the AC coder depends on the table descriptor, not on the component itself, and that some of the test streams use differing table descriptors, while other streams use the same table descriptors. Sorry folks, this one is definitely my fault and goes on my bill. So everything should be in proper order. So long, Thomas