Can quantization be done in HSV space for a CBIR system?If yes,what are your arguments?

I am having an argument with someone who said that I cannot do quantization in HSV space because this is a cylindrical space and I can't compute distances in a cylindrical space. Still, my cbir system returned nice results by using hsv quantification and the internet is full of the ' hsv quantification' topic which makes me think it is possible to do quantification in HSV space. But I didn't find any arguments to sustain why HSV quantification is possible or under what conditions and I ask you guys if you can give me any of these arguments and conditions.

Answer 1

[Walter Roberson]

For cylindrical coordinates, you are inside (or at most on the surface of) a convex polyhedra, and are allowed to travel through the polyhedra to reach points. A straight line between any two points in a convex polyhedra stays within the convex polyhedra (by definition of "convex"), so you can simply calculate Euclidean distance, same as if you were dealing with a cube.

Now if you were dealing only with the outside of a cylinder, the calculations would be a bit different, but that doesn't mean it cannot be done. After all, we deal all the time with distance calculations on the outside of a sphere (that is, Earth)

Answer 2

[Image Analyst]

What they said it not true, or else you didn't understand what they said. You certainly can quantize values, and you certainly can compute Delta E color difference in cylindrical spaces such as HSV or LCH. In fact, here is a slide from my color course that shows the formula for DE94:

You can see it's a complicated formula, and the updated DE00 (Delta E 2000) is even more complicated, but it can be done. Perhaps the person meant that you can't simply do deltaE=sqrt(deltaL^2+deltaC^2+deltaH^2), and that's true - it is a much more complicated formula.

If they need a more authoritative reference, tell them to look up this standard: ASTM D 2244-02 (or later one if available).