A Novel Technique Based on Texture Filtering for Medical Images

Abstract

ABSTRACT newlineBrain tumor is one of the major causes for the increase in mortality among children and adults. Magnetic Resonance Imaging (MRI) is a powerful visualization technique that allows acquiring images of internal anatomy in a safe and non-invasive way. Image segmentation is an indispensable process in the visualization of human tissues, particularly during clinical analysis of magnetic resonance (MR) images. It is the process of dividing the image into disjoint homogenous regions or classes. A key problem in medical imaging is automatically segmenting an image into its constituent regions. newlineRegion growing is a simple region based image segmentation algorithm. In normal region growing algorithm, the quotintensityquot constraint of neighbouring pixels is alone examined for growing the region. The normal region growing technique has two drawbacks: (1) noise or intensity variation leads to over segmentation and (2) shading of the real image may not be distinguished. In this thesis texture based region growing segmentation algorithm have been proposed. Two innovative modifications are done. In the first modification: texture and intensity based modified region growing (TIBMRG), the decision of growing the region into next pixel is based on pixel intensity and texture image (texture image is obtained by doing the LBP operator over the image). In the second modification: texture, intensity and orientation based modified region growing (TIOBMRG), the decision about assigning segment label is based on pixel intensity, orientation (which is obtained by applying gradient operator to the image) and texture image. Further in order to increase the performance (accuracy), it is optimized using genetic algorithm and is named as optimized region growing (ORGW) segmentation algorithm. newlineAt first the film artifact and noise are removed and the image is enhanced using Gaussian filtering and normalization. Secondly, features are extracted from the image and it is classified using KNN classifier. In the third phase optimized region growing