期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2014
卷号：69
期号：1
出版社：Journal of Theoretical and Applied
摘要：Manual segmentation by individual specialists on medical image dataset is time-consuming, expensive, and suffers from considerable inter and intra rater inconsistency. In addition segmentation is hard for the individual expert to combine the information from numerous portions and various channels when multi spectral data has to be examined. Unsupervised segmentation images as occlusions of textures, designed based on local histogram is well-suited to a broader class of images. The model proved that the local histograms were approximately the convex combinations of the value distributions of its component textures but did not provide with a richer characterization of textures and the pixel wise labeling consumed more time. Texture classification of images with multinomial latent model used a mixture density to obtain spatially smooth class segment. But better segmentation was not achieved for speckle noisy biomedical images and the texture classification of images increased the computational cost. To overcome the poor categorization of texture on medical images, the incorporation of neighborhood-based segmentation and binomial classifier tree-based sorting (NS-BCTS) is applied to demonstrate its utility in detecting the noisy speckle biomedical images in medical imagery. To start with, the neighborhood-based segmentation displays the features of rich set in terms of shape, position, color and neighborhood relations. The features extracted are then given as input to the binomial classifier tree-based sorting, with the data label obtained from the experts to minimize the time consuming process. The binomial classifier tree-based sorting examines each collective feature and labels it across the range to determine the computational cost. The experiment is conducted on biomedical image (i.e.,) lung cancer dataset with the factors such as time consumption, computational cost, running time, accuracy and feature categorization efficiency.
关键词：Segmentation; Neighbourhood-based segmentation; Binomial classifier tree-based sorting; Feature Categorization; Local value histograms; Medical imagery.
