Why convert to greyscale?

edit

We are doing an opencv study right now.

We are trying to match art paintings on mobile.

Now I'm looking for arguments to why or why not convert the images to greyscale first.

I think recognition could be improved keeping the color but since it's three layers instead of one it slows the progress down.

Everywhere on the internet I see examples who are seriously aiming at "the right way to do" converting the images to greyscale but nobody seems to be argumenting this.

So that's why I question this here:

Why should one convert color images to greyscale, what are the pro's and cons's?