RGB to DKL color space

Has anyone done any related work regarding the conversion of the RGB to DKL color space?

the only source i've found was http://search.googlecode.com/svn-history/r79/trunk/CG/matlab/simpsal/rgb2dkl.m, but it doesnt seem to be working correctly.

If anyone has a working script/function for this type of conversion (since the majority of color space conversions are already available to use in the database) plz give your input concerning this.

Thank you

[Walter Roberson]

"it doesn't seem to be working correctly": If you are going to expect us to play "20 Questions" to find out what problem you are seeing, then you are at least going to have to tell us the traditional, "Animal, Vegetable or Mineral" category to get us started.

Answer 1

[Walter Roberson]

The reverse of this, DKL to RGB, has python source at http://www.psychopy.org/epydoc/psychopy.misc-pysrc.html#dkl2rgb

Possibly the book http://spie.org/x648.html?product_id=881172 would have more information.

Note that DKL requires a calibrated monitor!

[VisLab]

thanks. I'll also give a look at that book, looks nice :)

Answer 2

[frank]

DKL is a color space which was used to model the colors in the LGN(visual system) whereas the standard Lab is more close to what happens in the Visual cortex. Since there is no processing happening between LGN and V1 hence ideally the signals should be the same yet they were observed to be different, hence dkl was defined which is slightly different from lab. One would find a big difference visually when comparing them. If somebody needs a code, you can go to Laurent Itti's lab webpage and download the visual saliency code which has a function rgb2dkl in matlab. I cannot upload the code since its there copyright, but its freely downloadable.Here's the link: http://www.klab.caltech.edu/~harel/share/gbvs.php

Hope this helps!!

[Walter Roberson]

The klab site does not appear to be available anymore.

I find one MATLAB conversion at https://github.com/sergeyk/vislab/blob/master/matlab/gbvs/util/rgb2dkl.m

Answer 3

[VisLab]

anyone ?

Answer 4

[Image Analyst]

No, I've never heard of that colorspace. What's it reason for being? Why can't you do what you need to do in a more standard space such as HSV or LAB?

[VisLab]

I'm looking for that colorspace so i can check if it gives better results than Lab or Luv colorspaces for Hue constancy.

If you know any other color-opponent colorspace besides Lab or Luv that you can recomend, please do so, i'll appreciate alot :)

[Image Analyst]

Sorry, but I've never worked with that colorspace. As far as I know CIE still recommends use of the CIELAB colorspace and Delta E 2010. I'm not sure what Hue constancy is. I know what color constancy is but not sure why that would depend on what colorspace you're using. Maybe you can try the newsgroup sci.engr.color.

[VisLab]

can u send me a link regarding Delta E 2010 please? I can only find CIE2000 dE.

Thank you

[Image Analyst]

It looks like I remembered incorrectly. I think the 2000 one is the latest, at least according to http://en.wikipedia.org/wiki/Color_difference Be aware though that dE is only valid for values less than about 5, although I use it for values up to about 20 for technical tests that don't need to be correlated that well back to humans.