Re: [Gimp-developer] OT: noise equivalence in HSV components
Joseph Heled writes: I might be wrong, but this will not solve it either. I assume the AB (color information) might be in the same units, but I don't think Luminosity is. (and I can make the argument that Hue and Saturation are in the same units as well, the problem being how to fit Value in) Well, I don't claim to be an expert on any of this stuff, but please take a look at http://www.dcs.ed.ac.uk/home/mxr/gfx/faqs/colorconv.faq Best, -- Bill __ __ __ __ Sent via the KillerWebMail system at primate.ucdavis.edu ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] OT: noise equivalence in HSV components
HSV is the wrong colorspace to use for this purpose. The LA*B* colorspace was designed to do what you are trying to accomplish: supposedly, equal distances in LA*B* coordinate space correspond to equal distances in human perceptual space -- although I understand that there is debate about whether this is really true. I might be wrong, but this will not solve it either. I assume the AB (color information) might be in the same units, but I don't think Luminosity is. (and I can make the argument that Hue and Saturation are in the same units as well, the problem being how to fit Value in) -Joseph ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
[Gimp-developer] OT: noise equivalence in HSV components
This is not a gimp question, but perhaps there is someone here who can shed light on this issue, How (or can you) combine errors/noise in HSV into one error/noise figure which reflects the total human visual error perception. I am sure this is not a good formulation of the question. Here is another way to put it, Take a picture I. Decompose to HSV. Now add noise Nh to H and recompose to get image I(h). Do the same for Ns to S and Nv to V. Now, what should the relation be between Nh,Ns,Nv so that a human would say All three images I(h). I(s) and I(v) look like they had the same amount of corruption relative to the original I. Now I am not necessarily asking for a rigorous answer. Any reasonable heuristic would do. -Joseph ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
Re: [Gimp-developer] OT: noise equivalence in HSV components
Joseph Heled writes: How (or can you) combine errors/noise in HSV into one error/noise figure which reflects the total human visual error perception. HSV is the wrong colorspace to use for this purpose. The LA*B* colorspace was designed to do what you are trying to accomplish: supposedly, equal distances in LA*B* coordinate space correspond to equal distances in human perceptual space -- although I understand that there is debate about whether this is really true. (Of course, you have to be careful about what you are trying to do. If you shift an entire image one pixel to the right, it will probably look identical to a human although it may be quite different on a pixel-for-pixel basis. In general, integrating pixel-by-pixel color differences will not give you a very accurate picture of how much difference people perceive between two images.) The decompose plug-in in Gimp will decompose an image into LAB coordinates -- however, I looked at the source code recently in connection with a bug report (bug #147603), and it seems that the LA*B* algorithm is not correctly implemented by the plug-in, so you should be cautious in using it for anything important. Best, -- Bill __ __ __ __ Sent via the KillerWebMail system at primate.ucdavis.edu ___ Gimp-developer mailing list [EMAIL PROTECTED] http://lists.xcf.berkeley.edu/mailman/listinfo/gimp-developer
