Research On Interpolation Of Medical Image And Image Segmentation In 3D Reconstruction System

The main work of this paper are the interpolation of image between two neighboring slices image and segmentation technology study in medical image 3D visualization and the realization of medical image 3D reconstruction. Medical image 3D visualization is to use the 2D images information to reconstruct 3D image, then stereoscopic display, and conduct various kinds of medical image data processing.Commonly, the distance between two neighboring slices is large. In order to get more details, an interpolated slice is needed. Traditional method of point-to-point registration has low registration precision. In this paper, we present a non-rigid registration based on B-splines which optimizes the normalized mutual information similarity measure, which is proposed to yield the deformation between adjacent slices. Compared with traditional method, new method improved registration precision of whole image. A modified bilinear interpolation method is then designed to generate propagating image by finding a minimal surrounding quadrangle.Medical image segmentation is to divide the medical image into several regions, and then extract the interested regions of tissues or organs. Segmentation is an essential step in the medical image processing, the accuracy of segmentation is very important for the follow-up medical image processing and the doctor to assess the real situation of the diseases. Because of the complexity of medical images, so far there is not any all-purpose segmentation method. In this paper, we present a watershed segmentation method with haralick's Texture-Based Region Merging. The original image was preprocessed with Gaussian filter and Sobel filter, on which thresholding is performed to remove weak edge pixels. The normalized mutual information similarity measure we present as evaluation function applies to the similarity in textural features belonging to two neighboring partitions.