A robust Auto Calibration technique for Stereo Camera

To get accurate image results, the camera should be calibrated accurately as well. There are several methods and techniques of calibration that have been proven to effectively work but this paper specifically focuses on the robust approach to Auto Calibration of stereo camera. On this, the paper only exploits the geometric constraint, namely, the epipolar geometry and calculation. The paper also gives an overview of the method for Comparison; procedures and the related literature in the topic of auto calibration.

  1. 1


    Calibration of cameras is very important for any application that requires establishment of geometrical relation between images and specific scenes. Such applications include the stereo image rectification and tracking of objects.
    Basically, calibration of cameras is essential in the following ways;
    a) Used to estimate the intrinsic parameters of the camera, for instance the focal length, principal point, aspect ratio, skew, and lens distortion
    b) Used in estimating the extrinsic parameters of the camera such as the camera poses in the 3D space
    Conventionally, there are three methods that have been used over time with relatively good results; the single camera calibration, multiple camera calibration and the self or auto calibration methods. Single camera calibration has seen various developments and has been the basis for some of the recent most popular techniques. For instance, the Zhang’s Methods which are suitable for video sequences with at least several hundreds of frames and requires at least two different views for better precision. Another method that is very common in single calibration is the Faugeras and Toscani technique which requires the availability of a non-planar calibration pattern and is commonly used in estimating the PPM when the lens distortions are not modeled. Both of these methods have been proven to lack accuracy and are very tedious to use [1].
    Although multiple-camera calibration methods offer solutions to the challenges related to the image acquisition process which is often very tedious in single calibration, the method cannot be used to calibrate only one camera. Its accuracy is also dependent on the number of cameras used [1]. On the other hand, the recent experiments indicate that the proposed robust approach produce better results compared to the existing techniques even from a single image calibration [1-4].

  2. 2

    The Procedure

    The robust approach of auto calibration employs seven techniques for Feature Extraction which include; SURF, BRISK, FAST, FREAK, MinEigen, MSERF and SIFT and also tested [5]. A comparative analysis is normally done to identify the perfect method. One the method is identified, correspondence is established between points extracted in stereo images with Various Matching Techniques (SSD, SAD, and Hamming). Fundamental Matrix is then calculated which helps to estimate the epipolar line by choosing the perfect Eight-point algorithms (Norm8Point, LMedS, RANSAC, MSAC, and LTS) [6].

  3. 3

    Related work

    There exist several methods and procedures of camera calibration. The key aspects of consideration are; the ease of use especially for the common user and the precision that enable accurate constructions. Generally, some methods may be easier to perform but lack accuracy and it is worth noting that even a small calibration error can result into large coordinate errors which greatly impact the precision of the whole system.
    There have been positive developments on the topic of camera calibration over the last few decades. In an effort to improve on both of the aspects precision and ease of use, past studies have proposed various methods and procedures. For instance, Moreno and Taubin [2] proposed a new method that he found simpler and more accurate and which employs full pinhole model that include the radial and tangential lens distortions in describing camera behaviors.
    To solve the issue of tediousness common in single camera calibration techniques, Caprile and Torre [7] proposed a multiple camera calibration technique that exploits the properties of the vanishing points. The method enables the determination of rotation matrix and translation vectors using the intrinsic and extrinsic parameters of the images. This method however does not estimate the distortions and thus leaves room for lack of accuracy.
    In consideration, standard cameras and projectors uses checkerboards that is manually operated when determining different calibration parameters [Deglint, et al. [1], 8]. Calibration of cameras when done manually is generally costly, time consuming and error prone. To address these concerns, the use an auto-calibration technique is advisable. It offers accurate images, is less tedious and is very useful in building 3D models and enables geometric screen corrections [1].
    Elsewhere, Vupparaboina, et al. [9] propose the use of Euclidean auto calibration which recovers both shape and scale; the features that are very essential in applications such as multi-party telepresence. The method also offers great accuracy is less tedious compared to Zhang and Pollefeys methods [10].

  4. 4


    It is worth noting that advancement in research and technology has delivered positive benefits in every sector of economy. Such advancements have been witnessed in the field of calibration and have led to development of various methods that are aimed at providing solutions to the challenges of the old manual methods. Such developments include the robust approach of auto calibration which has been proven to be both accurate and easier to use compared to the previous methods such as Zhang and Faugeras and Toscani techniques.

  5. 5


    [1] J. Deglint, A. Cameron, C. Scharfenberger, H. Sekkati, M. Lamm, A. Wong, et al., "Auto‐calibration of a projector–camera stereo system for projection mapping," Journal of the Society for Information Display, vol. 24, pp. 510-520, 2016.
    [2] D. Moreno and G. Taubin, "Simple, accurate, and robust projector-camera calibration," in 3D Imaging, Modeling, Processing, Visualization and Transmission (3DIMPVT), 2012 Second International Conference on, 2012, pp. 464-471.
    [3] A. Gruen and T. S. Huang, Calibration and orientation of cameras in computer vision vol. 34: Springer Science & Business Media, 2013.
    [4] Z. Zhang, "A flexible new technique for camera calibration," IEEE Transactions on pattern analysis and machine intelligence, vol. 22, pp. 1330-1334, 2000.
    [5] S. Nazir, A. Assoum, B. El Hassan, and F. Dornaika, "Augmented Reality application: A new implementation chain," in Multidisciplinary Conference on Engineering Technology (IMCET), IEEE International, 2016, pp. 69-74.
    [6] A. A. Fahmy, "Stereo vision based depth estimation algorithm in uncalibrated rectification," Int J Video Image Process Netw Secur, vol. 13, pp. 1-8, 2013.
    [7] B. Caprile and V. Torre, "Using vanishing points for camera calibration," International journal of computer vision, vol. 4, pp. 127-139, 1990.
    [8] S. Placht, P. Fürsattel, E. A. Mengue, H. Hofmann, C. Schaller, M. Balda, et al., "Rochade: Robust checkerboard advanced detection for camera calibration," in European Conference on Computer Vision, 2014, pp. 766-779.
    [9] K. K. Vupparaboina, K. Raghavan, and S. Jana, "Euclidean auto calibration of camera networks: Baseline constraint removes scale ambiguity," in Communication (NCC), 2016 Twenty Second National Conference on, 2016, pp. 1-6.
    [10] G. Hua, Y. Fu, M. Turk, M. Pollefeys, and Z. Zhang, "Introduction to the special issue on mobile vision," International Journal of Computer Vision, vol. 96, pp. 277-279, 2012.

Don't like this list? Edit it and make your own list!

Don't like this list? Edit it and make your own list! We will pubish it on our site! You can share it with your friends on Facebook, Twitter, etc

Edit this list

Login / Sign up