The main goal of this master’s thesis was get up used methods for observation the stereoscopic scene with one couple of cameras and find out good solving for processing these resulting pictures for two-view and multiple-view autostereoscopic displays for three-dimensional perception. For methods for acquisition video was introduced two methods. They were method “Off-axis” with parallel camera axis and method “Toe in” with intersections axis. For fit method was choice the method named as “Off-axis“. It was not produces the vertical parallax and in detail was in this work described principle of this method. Further were describe principles off used methods for three-dimensional perception namely from the oldest method named anaglyph after methods for viewing at autostereoscopic displays. The Autostereoscopic displays were main goal of this thesis and so their principles were described in details. For production the result image for autostereoscopic displays was used generation intermediate images between left and right camera. Resulting videos were acquisition for testing scene in created in 3D studio Blender, where was possible setting system of cameras exactly parallel axis. Then were introduce principles processing video where was extract from the couple of cameras where were connected to PC for help digitizing card and next time with two web cameras. Here is not guaranteed exact parallel axis system. Therefore this work try for real cameras achieve exactly parallel axis system by the help of transformations of frames with stereo rectification. Stereo rectification was solving with OpenCV libraries and was used two methods. Both methods work from principles epipolar geometry that was described in this work also in detail. First method rectifies pictures on the basis fundamental matrix and found correspondences points in two images of the scene and second method rectifies pictures from knowledge intrinsic and extrinsic parameters of stereoscopic system of cameras. In the end of this work was described application for implementation introduced methods.

The ability to accurately map and localize relevant objects surrounding a vehicle is an important task for autonomous vehicle systems. Currently, many of the environmental mapping approaches rely on the expensive LiDAR sensor. Researchers have been attempting to transition to cheaper sensors like the camera, but so far, the mapping accuracy of single-camera and dual-camera systems has not matched the accuracy of LiDAR systems. This thesis examines depth estimation algorithms and camera configurations of a triple-camera system to determine if sensor data from an additional perspective will improve the accuracy of camera-based systems. Using a synthetic dataset, the performance of a selection of stereo depth estimation algorithms is compared to the performance of two triple-camera depth estimation algorithms: disparity fusion and cost fusion. The cost fusion algorithm in both a multi-baseline and multi-axis triple-camera configuration outperformed the environmental mapping accuracy of non-CNN algorithms in a two-camera configuration.

This thesis deals with the task of three-dimensional scene reconstruction using image data obtained from multiple views. It is assumed that intrinsic parameters of the utilized cameras are known. The theoretical chapters describe the basic priciples of individual reconstruction steps. Variuous possible implementaions of data model suitable for this task are also described. The practical part also includes a comparison of false keypoint correspondence filtering, implementation of polar stereo rectification and comparison of disparity map calculation methods that are bundled with the OpenCV library. In the final portion of the thesis, examples of reconstructed 3D models are presented and discussed.