Segmenting Medical Image Data Set with Inhomogeneous Intensities by Using Level Sets

Abstract

The images obtained during the medical procedure have inhomogeneous intensity properties which cause the difficulty in image segmentation. The globally accepted region-based algorithm relies on the homogenous intensity in interest regions. So, the expected result cannot be obtained for medical images. In this research work, we induce a functionality criterion (FC) to cluster the locals (CL) according to the intensity of medical images. The FC is merged with the mid of neighbor to provide an accepted criterion for image segmenting. By formulating level sets, the functionality level set criteria (FLSC) has an energy that represents the domain portion of the image. By reducing this energy, the medical images can be consecutively segmented. The validation of our method has been done on 100 datasets of endoscopic medical images with inhomogeneous intensities. The results show that the proposed technique provides promising results and high performance than well-known algorithms.

 Keywords: Image segmentation, intense pixels, weight cut, averaging cut, normalization cut.