الفهرس 
Chapter 1 IntroductionOnly 14 pages are availabe for public view
 |  | from | 127 |  |  |
| | | from | 127 | | |
Abstract
Recovering the 3D structure of a scene together with the camera motion from a sequence of 2D images instead of a single image is known as Structure from Motion (SFM) problem. It is one of the widely researched problems over the last two decades. Features occlusion is one of the major problems with SFM, which comes from the panoramic camera movement around the object. It refers to the appearing and disappearing of feature points throughout the camera movement. Many research works have been proposed to tackle the problem of features occlusion to reduce the computation time, which results from estimating the occluded feature points before reconstruction. In all cases they still try to estimate the missing 2D points before applying the SFM algorithm. In this thesis, a fast method for 3D shape reconstruction from video sequence is proposed. The proposed method is based on registering multiple partial reconstructions (patches). The method tackles the problem of features occlusion in a different manner without the need to estimate the missing 2D feature points in all images. In addition, a texture mapping pipeline is proposed to produce a consistent 3D textured surface. This thesis is organized as follows. Chapter 1 gives an introduction to 3D reconstruction techniques, motivation, problem statement and objectives. Chapter 2 gives an overview of the SFM techniques and related work. Feature detection and tracking is introduced in chapter 3. It is an important step in any 3D reconstruction algorithm. Chapter 4 explains one of the most successful SFM methods, the factorization method. While in chapter 5, the proposed method of sub-matrices selection to handle the feature occlusion is explored in details. The proposed texture mapping pipeline is introduced in chapter 6 to provide a realistic 3D textured shape. vi Experimental results on both synthetic and real data are presented in chapter 7. Chapter 8 concludes the work done and points to some future ideas in the field.