What do you think?
Rate this book
3D Object Reconstruction from Stereo Images
3-D reconstruction of a scene from its 2-D images captured from different viewpoints is one of the fundamental problems of computer vision. It is an active research area since the last three decades and has many applications like automatic navigation of mobile robots, automatic vehicle driving, feature and object identification, creation of 3-D maps and terrain mapping from satellite imaging. Further, it can be useful in science, engineering and medicine applications where human vision-like capability is required. The theoretical fundamental of the 3-D reconstruction is well established and is based on triangulation principle. Fundamentally, stereo vision involves two the fusion of features observed by two cameras and the reconstruction of their three dimensional model. When a single image feature is observed by the cameras, stereo vision is an easy problem. However, image of a scene consists of thousands of pixels with thousands of image features such as edge elements, similar texture, etc. In this situation stereo vision is one of the most difficult research problems of machine vision and eluding researchers for past three decades.
The primary issues for 3-D reconstruction of a scene are camera calibration, correspondence problem, occlusion handling and 3-D reconstruction. Among these the most important one is the correspondence problem that refers to finding out which item in the first image corresponds to which item in the second image. The displacement between two locations in the left and the right images of a same scene point is called as disparity and it is inversely proportional to depth in the scene. The disparity of all the image points is called disparity map and is obtained by identifying corresponding points in both images that are the projections of the same point in the scene. However, assigning each point to its correct disparity is a challenging task particularly in textureless areas and at depth discontinuity. Occlusions, features visible to only one camera, have further compounded the correspondence matching as occlusions occur at object boundary.
The last three decades of research in stereo correspondence matching has resulted in a large number of techniques that can be classified as local methods, global methods and segmentation based methods. Each method has its own merits and demerits. For example local methods have potential for real time generation of disparity map but do not produce good results as compared to global methods. Global methods produce better result but are computationally expensive. Segmentation based methods first segment the images and then use global methods or local methods for disparity computation. Presently, no method finds the disparity map accurate enough and in real-time to meet the requirement of real applications. Hence, further research is required to make it suitable for actual applications.
The book is outcome of the doctoral work of the author and is primarily focused on to introduce the basic concepts of stereo vision along with the recent development in the field. The book consists of six chapters. The first chapter introduces the basic concepts of a camera system for imaging along with the fundamental concepts of stereo vision. It also highlights various research issues in the field. Chapter 2 discusses the basic concepts of global and local correspondence matching techniques. Chapter 3 discusses the implementation aspect of few window-based correspondence methods and discusses related research issues. Chapter 4 discusses the fundamental of color segmentation based correspondence matching algorithms. Chapter 5 introduces the use of dynamic programming to solve correspondence matching. Finally, chapter 6 discusses the experimental setup used to capture stereo images and discusses the 3D reconstruction process. It also discusses various open research issues that need further research to solve actual problems.
The primary issues for 3-D reconstruction of a scene are camera calibration, correspondence problem, occlusion handling and 3-D reconstruction. Among these the most important one is the correspondence problem that refers to finding out which item in the first image corresponds to which item in the second image. The displacement between two locations in the left and the right images of a same scene point is called as disparity and it is inversely proportional to depth in the scene. The disparity of all the image points is called disparity map and is obtained by identifying corresponding points in both images that are the projections of the same point in the scene. However, assigning each point to its correct disparity is a challenging task particularly in textureless areas and at depth discontinuity. Occlusions, features visible to only one camera, have further compounded the correspondence matching as occlusions occur at object boundary.
The last three decades of research in stereo correspondence matching has resulted in a large number of techniques that can be classified as local methods, global methods and segmentation based methods. Each method has its own merits and demerits. For example local methods have potential for real time generation of disparity map but do not produce good results as compared to global methods. Global methods produce better result but are computationally expensive. Segmentation based methods first segment the images and then use global methods or local methods for disparity computation. Presently, no method finds the disparity map accurate enough and in real-time to meet the requirement of real applications. Hence, further research is required to make it suitable for actual applications.
The book is outcome of the doctoral work of the author and is primarily focused on to introduce the basic concepts of stereo vision along with the recent development in the field. The book consists of six chapters. The first chapter introduces the basic concepts of a camera system for imaging along with the fundamental concepts of stereo vision. It also highlights various research issues in the field. Chapter 2 discusses the basic concepts of global and local correspondence matching techniques. Chapter 3 discusses the implementation aspect of few window-based correspondence methods and discusses related research issues. Chapter 4 discusses the fundamental of color segmentation based correspondence matching algorithms. Chapter 5 introduces the use of dynamic programming to solve correspondence matching. Finally, chapter 6 discusses the experimental setup used to capture stereo images and discusses the 3D reconstruction process. It also discusses various open research issues that need further research to solve actual problems.
111 pages, Paperback
Published November 29, 2017
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
No one has reviewed this book yet.