ABSTRACT
The research provides a review of segmentation methods for medical imaging. The article provides a comparative study of edge-based, region-based and energy-based methods of image segmentation. Many medical images have different levels of intensity because of flaws in the object or technical limitations. Some segmentation techniques have limitations, like being stuck in local minima or producing over-segmented images. Medical image segmentation is still difficult because of noise, poor contrast, and huge variations in intensity level. Interactive medical image segmentation is also needed for better results that can be solved by taking user input into account while segmenting an image. Active contour techniques assume that global intensity may be used to define an image. Some approaches deal with intensity inhomogeneity in the same way that the region-scalable fitting (RSF) model does. The method was compared to Chan-Vese (CV), RSF, and local and global intensity fitting (LGIF). The hybrid region-based active contour (HRBAC) approach may be useful in addressing the intensity of inhomogeneity. It also accelerates segmentation as compared to local region-based active contour (LRBAC). However, HRBAC has limitation of contour initialization sensitivity and parameter sensitivity. Improved HRBAC model is also explored in this paper to deal with challenges of medical image segmentation mentioned above. The new function used in this model will leverage global and local data to quickly get the correct response. The method combines energy functional-driven curve generation with a level set framework for medical picture segmentation. Lattice Boltzmann approach is used to make segmentation process fast.
