Tumor demarcation of MRI scan using KMCG vector quantization algorithm
Hemant B. Kekre, Tanuja Sarode, Saylee Gharge, K. Raut · 2011
In this paper, texture based segmentation algorithms are considered for comparison. The problem with some of these methods is, they need human interaction for accurate and reliable segmentation. Human interaction is in terms of providing some initial knowledge externally for segmentation. This knowledge is in terms of a small amount of labeled data for some or all classes. This is usually time-consuming and expensive. Segmentation based on Gray level co-occurrence matrix gives better result for variance but computational complexity is more. Watershed gives over segmentation. Morphological provides better segmentation but edges can not get eliminated whereas for segmentation using Kekre's Median Codebook Generation (KMCG) algorithm shows proper tumor demarcation by avoiding other part of the image.