I am working on intelligent color quantization (reduction), for this problem I think we can consider a color space in which, the color data can be quantized into groups of major or dominant color with less error.
Any recommendation is invited...
Thanks a lot, if I study that part I can get idea and learn how to propose a new color space model? And use it for color reduction or other image processing purpose.
If so, may you refer a link or website to your book chapter.
I agree , color space(s) are standardized and used in a lot of algorithms (jpeg and mpeg) and printers (CMKY) so looking and optimizing the # of channels will lead to the same set. Right now it is shown 2 channels are enough to generate a whole range of color(s) .
Quantization there you can look into optimization/adaptive, depending on the color coordinate system to reproduce the true color .
http://en.wikipedia.org/wiki/CIE_1931_color_space
Refer Video Demystified by Keith Jack
I agree with the the other answers. Inventing a new color space would be pointless for any application I could think of. However, quantizing the existing color spaces for human experience would be a most welcome effort. The MacAdam ellipsoids are so few and far apart. But you have to remember that the number of discernible colors are still estimated to be in the millions.
You can look at illuminant invariant image generation process- this will give you an idea about inventing a new color space to match your needs. Here's a link:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.15.4223&rep=rep1&type=pdf
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing