What is non-maximal suppression?

In the harris corner detector code a few lines from the bottom he performs non-maximal suppression. He has an explanation of what it is doing, but I don't understand it fully. Can someone explain what it is doing and if it's important or not?

Answer 1

[Daniel Costa]

Think about the value of R for all the points in a patch that contains a corner: it is likely that more than one cell > threshold (intuition: moving the window by one pixel will not change the image much). Thus for each candidate point > threshold, we detect if it is the absolute maximum within its territory, if not then we discard it.