Structured light systems are an important tool for reconstructing 3D geometric details of objects lacking in texture, where finding correspondences between images using passive stereo is difficult. Catadioptric vision systems employ curved mirrors and different types of lenses to broaden the field of view as compared to a conventional camera lens, allowing a larger portion of the overall scene to be imaged in every shot. Although there have been some research papers published in these areas in the recent years (mostly by the industrial automation community), almost no other recent book focusing specifically on this subject can be found.
This is a small book (155 pages) providing a good literature review and discussing the work of the authors on some problems in active vision systems and catadioptric systems. These problems include static internal calibration of a camera, relative orientation estimation of the camera with respect to the projector when it is allowed to move, 3D reconstruction using an image-to-world transformation when certain geometric properties about the scene are known, and modeling, calibration, even mechanical alignment of different catadioptric systems. Although it is easy to follow, I have not fully checked the mathematics for technical correctness, however, given that this is mostly based on work published in different journals, I am assuming that the techniques have already been independently checked and found to be correct.
The first chapter provides a very useful and detailed literature survey on work done in different problems of geometric computer vision including calibrated and uncalibrated 3D reconstruction, giving special attention to structured light systems. This literature survey can be very helpful for any practitioner or student interested in finding out what the problems in geometric vision have been, and how progress in those problems has been made over the previous few decades.
The second chapter defines important fundamental concepts in structured light systems and catadioptric systems. The purpose of this chapter is to make the book self-contained, so that all the concepts used in the next four chapters, which form the ‘meat’ of the book, are available. I don’t think the chapter succeeds in meeting this objective, because on one hand, it describes some basic concepts like fundamental matrices, planar homographies, cross-ratios, but on the other hand, it does not discuss other equally fundamental ideas like vanishing lines, conics, line at infinity, etc. that are also used in the coming chapters. Thus, if a reader is not aware of basic ideas in geometric vision, they will need to refer to a standard textbook, which, for students, reduces the advantage of reading this book instead of going through the published papers.
The third chapter describes the determination of internal camera parameters based on three different planar patterns as well as a method for nonlinear distortion correction. These patterns include a polygon whose vertices lie on a circle, on two intersecting circles, and on two concentric circles. Although the pros and cons of these planar patterns have been discussed individually, no experimental comparison between them is performed, and experimental validation (numerical simulations and limited real experiments) is only given for the last method, which seems comparable to state-of-the-art. In addition, another planar pattern based method for distortion correction is discussed. However, no experimental evaluation or comments on how this method compares to other methods in the literature is provided.
Estimation of relative orientation if the camera is allowed to move with respect to the projector is discussed in the fourth chapter. This approach requires that an arbitrary planar object be present in the scene so a homography can be computed between the original and the projected pattern. The method of the authors assumes that the internal calibration parameters of the camera are known, and does not guarantee a unique solution for the relative orientation. The chapter also performs error analysis and shows results of experiments on simulated and real-world settings. It is well-written and the mathematics is easy to follow.
The fifth chapter describes an interesting approach to 3D Euclidean reconstruction by estimating an image-to-world transformation matrix given that there are one or two known planes in the scene. The projected colored pattern is treated as a collection of light planes each comprising a line in the pattern together with the projector center. The intersection of these light planes with the scene is used to estimate the 3D geometry. However, the system is based on colored patterns, and its performance with objects that are also colored (thus causing a wavelength shift when the light falls on them) and well-textured is unreliable.
Catadioptric vision using central projection is considered in the sixth chapter, discussing in detail three different systems: a hyperbolic mirror and fisheye lens camera, a parabolic mirror and an orthographic camera, and another comprising a hyperbolic mirror with a perspective camera. The chapter begins with a nice overview of different techniques used for extending field-of-view and then moves on to discuss fine details relating to the mechanical installation of the components in such systems as well as system modeling, different calibration procedures, and 3D reconstruction. It also discusses comparative pros and cons of these systems. Overall, the chapter provides interesting insights and can be recommended.
The last chapter gives a brief conclusion of the whole book and then discusses ideas for future work in the problem domains covered previously. Specifically, these include suggestions for image-to-world transformation based 3D reconstruction when nothing about the scene structure is known apriori, integrating structured light techniques with shape from shading, enabling color-encoded light patterns to work in strongly colored scenes, and catadioptric structured light systems.
Overall, the literature survey provided by the book on different problems is a valuable one, and also the example methods for different problems related to reconstruction using active and catadioptric vision systems give a good outlook into the field. It can be worth collecting for practitioners as well as for students studying active or catadioptric vision systems.
Calibration and Reconstruction for Active Vision Systems
Beiwei Zhang and Y. F. Li
Intelligent Systems, Control and Automation: Science and Engineering
M. Zeeshan Zia (Switzerland)
Recent book reviews:
Multispectral Satellite Image Understanding
By Cem Ünsalan and Kim L. Boyer
Automatic Digital Document
Processing and Management:
Problems, Algorithms, and Techniques
By Stefano Ferilli
Automatic Calibration and Reconstruction for Active Vision Systems
By Beiwei Zhang and Y. F. Li
Fundamentals of Digital Image Processing:
A Practical Approach with Examples in Matlab
by Chris Solomon and Toby Breckon
An Introduction to Biometrics
by Anil Jain, Arun Ross,
and Karthik Nandhakumar
Handbook of Geometric Computing
by Eduardo Bayro Corrachano (Ed.)
Essential Image Processing and GIS for
by Jian Guo Liu and Philippa Mason
Handbook of Pattern Recognition and
Computer Vision, 4th Edition
by C.H. Chen