New Region Growing Segmentation Technique for MR Images with Weak Boundaries

摘要

In medical image processing, segmented images are used for studying anatomical structures, diagnosis and assisting in surgical planning. Our goal is to segment 2D/3D MR images which contain weak boundaries between different tissues. Region growing method cannot specifically segment the tissues with weak boundaries because pixels inside the region and outside have similar intensity. We introduce a new automatic threshold method for region growing. The pixels around boundaries have lower probability of intensities than the pixels inside and outside the tissues. Then, the automatic threshold for region growing will be computed as a function of pixel intensity value and probability of intensities. The experimental results show that the proposed technique produces accurate and stable results.