3D volumetric reconstruction from 2D images

Volumetric visualization requires higher processing load than surface visualization. In parallel with the development of technology, interest in volumetric visualization has been increased in recent years. One of the major challenges in volumetric visualization is the extraction of regions of interest through transfer functions. In this thesis, a classification-based approach, which is frequently used in surface visualization, has been applied to solve this challenge in volumetric visualization. In the approach proposed in this thesis, volumetric visualization process is carried out in two stages. In the first stage, local boundaries are extracted with the image processing and the region of interest (ROI) is determined by combining these boundaries with the machine learning methods. In the second stage, ROI is visualized volumetrically by using a transfer function. While analysing the performance of the proposed approach, the effect of the noise on the ROI and the volumetric image was investigated. As a result of the performance analysis, it was seen that the proposed approach improved the volumetric visualization.