Fast and Intelligent Determination of Image Segmentation Method Parameters

Advanced digital image segmentation framework implemented by using service oriented architecture is presented. The intelligence is not incorporated just in a segmentation method, which is controlled by 11 parameters, but mostly in a routine for easier parameters’ values determination. Three different approaches are implemented: 1) manual parameter value selection, 2) interactive step-by-step parameter value selection based on visual image content, and 3) fast and intelligent parameter value determination based on machine learning. Intelligence of second and third approach is introduced by end-users in the repeated interaction with our prototype in attempts to correctly segment out the structures from image. Fast and intelligent parameter determination predicts a new set of parameters’ values for current image being processed based on knowledge models constructed from previous successful (positive samples) and unsuccessful (negative samples) parameter selections. Such approach pointed out to be very efficient and fast, especially if we have many positive and negative samples in the learning set.