Kernel Density Estimation
•Assume p(I(x,y)=k) similar for similar values of k. •So observation of k tells us a new observation at or near k is more likely.
•Equivalent to smoothing distribution.
•(smoothing reduces noise)