Medical Image Segmentation Using a Robust Edge-stop Function with 2×2 Window Patch

Edge-based active contour models using robust edge-stop functions (ESFs) have shown their effectiveness in segmenting medical images. The ESFs utilize classification scores from machine learning algorithms which are sensitives to the feature vector. The number of features influences the speed of the algorithms. Previous studies utilize image patch 3×3 which consists of 9 components in the feature vector and leads to a long computational time. This paper investigates the use of a simple feature vector for the ESFs. An image patch of 2×2 is selected and applied to a classification algorithm, namely the k-nearest neighbors (k-NN). Experimental results indicate that the feature leads to similar in accuracy but faster in computational speed.