Abstract
In this thesis, a multiresolution method（or called pyramid-based method）in combination with fuzzy c-means clustering method （FCM）for 3D MRI segmentation is described. A pyramid of the image is first constructed and smoothed by using a low-pass filter. The pixel values of the second layer of the pyramid of the T1 and T2 images are viewed as the two axes of the feature plane. The fuzzy c-means algorithm is then applied to the feature plane. Each cluster represents one tissue. Finally, we transform the images back to the original resolution and compare the difference with the segmented original image based on the distance from the cluster center to sift the noise. To examine the cluster result, the uniform data functional （UDF）is used for measuring the quality of the FCM algorithm. It shows that the ideal number of the clusters is the same as the brain tissues. The segmentation result of this method is better then other FCM-based methods in terms of noise reduction.