| Author | Ngo Cao Dinh |
| Call Number | AIT Thesis no.CS-08-12 |
| Subject(s) | Three-dimensional imaging
|
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. CS-08-12 |
| Abstract | This thesis presents a system for constructing 3D models from CT images. The first thing I present is the general methodology and some fundamental techniques to develop a 3D application. To represent 3D models built by the system, I discuss surface and volume rendering. For surface rendering, I use the marching cube algorithm to generate a mesh of the contour in volumetric data based on the intensity values before projecting this surface onto the view plane. For volume rendering, I present two main kinds of techniques: image-order and object-order volume rendering. Relating to image-order volume rendering, I use ray casting algorithm to project the volume metric data onto the image plane. I also introduce some compositing functions used to calculate values of the pixels on the image plane during running this algorithm. The most important part of the 3D modeling system is to provide functions for filtering and segmentation. To implement those functions, I present some algorithms used to filter and segment 2D raw image data such as binary threshold filter, general threshold filter, canny edge detection filter, Gaussian smoothing filter, region growing based on connected threshold segmentation. I also present some algorithms applied to volumetric data loaded from the raw or preprocessed 2D images such as 3D Gaussian smoothing filter, order static filters, dilation filter, erosion filter and mesh smoothing filter. By comparing the result models generated by various segmentation pipelines, we determined that the teeth models generated by the segmentation pipeline with the median filter are the most precise and smooth and most suitable for simulation in dental training. The system has been tested with the CT scans of the teeth and jaw bone of different qualities. The results of this test show that it is robust enough to segment data sets containing both a little noise and large amounts of noise. The system also accepts data sets coming from different types of CT machines. Moreover, it is capable of segmenting arbitrary parts of the body. This generality was tested by segmenting a data set of the thigh. However, the maximum number of segmented tissues types supported by this system is limited to three. |
| Year | 2008 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. CS-08-12 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Haddawy, Peter |
| Examination Committee(s) | Dailey, Matthew;Siriwan Suebnukarn |
| Scholarship Donor(s) | Asian Institute of Technology Fellowship |
| Degree | Thesis (M.Sc.) - Asian Institute of Technology, 2008 |