This paper is a final solution for a two semester long diploma thesis. The goal was to develop a system that is able to reconstructe points from 2D images to the real spatial world as a point cloud. For the implementation of this complex algorithm, I had to do the following sub-tasks.
I need the intrinsic information of the camera that made the images, and efficient image processing and feature point extraction. Furthermore I needed to pair the images with their feature points, and to calculate the exact position of these coordinates and determine the relativ shift between the first and second cameras. With this process I was able to build a 3D point cloud, that led me to construct the surface reconstruction of the original spatial scene. This process then can be refined a number of ways, such as median filtering or using 3 image windows.
For the implementation I used an existing, open source, image processing library, the OpenCV. The results were evaluated on the basis of objective and subjective criteria, using 3D computer graphic programs.