The HST produced some deep field images in mostly visible light, which contain thousands of galaxies that range in color from white to red (e.g. see here). I understand that the interpretation of astrophotographic images is confused by image processing and colorization of IR and other light. But for deep field images of galaxies in the visible range, can I interpret the "red looking" galaxies as being red due to redshift? I would have assumed this to be the case if it were not for the fact that the redder galaxies do not seem obviously correlated with smaller angular diameter (i.e. they do not look farther away). Are galaxies naturally so diverse in color that it is a mistake to interpret redder galaxies as being redshifted?
Yes, there is considerable diversity in the intrinsic colours of galaxies. Anything from "blue" galaxies dominated by star formation burst with lots of hot, young stars, to old ellipticals dominated by red giant populations and old main sequence stars.
To get a redshift from the colour would generally need more than a red Vs blue judgement. You would need measurements through several coloured filters and even then there can be considerable ambiguity and uncertainty. These days, machine learning and artificial neural networks are deployed to make these estimations (D'Isanto & Polsterer 2017).