Paths Lengths in Stochastic Graph Image Pyramid1)

In many applications some parts of an image are of special interest. We present in this paper results of the analysis of the vertical path lengths in stochastic graph image pyramids. Such path lengths evaluate the efficiency of the pyramid structure needed e.g. for algorithms which derive object properties from the object pixels in the image. Our aim is to build stochastic image pyramid locally that are optimal in the sense of bottom-up and top-down processes.