Parameter setting through design of experiments for the automated delineation of regions of interest

In the field of image processing, the complete implementation of segmentation procedures, in order for example to delineate automatically regions of interest in images, requires usually adjusting various parameters, which interfere each with the others. Analysis of all the possible combinations is cumbersome and time consuming. Moreover, settings obtained should remain valid for families of images of the same type. In this paper, in view of the accurate delineation of an object out of gray level images, we describe a method for efficient parameter setting of a level set based segmentation procedure. The experimental approach proposed makes use of a Design of Experiments methodology in order to identify the most influential parameters and optimize their values. This allows the optimal determination of a mask delineating the object of interest in gray level images and necessary for the following application dependant processing steps.