Page 314 -
P. 314

6.3 Geometric intrinsic calibration                                                    293

























               Figure 6.11  Four images taken with a hand-held camera registered using a 3D rotation motion model, which
               can be used to estimate the focal length of the camera (Szeliski and Shum 1997) c   2000 ACM.



               6.3.4 Rotational motion

               When no calibration targets or known structures are available but you can rotate the camera
               around its front nodal point (or, equivalently, work in a large open environment where all ob-
               jects are distant), the camera can be calibrated from a set of overlapping images by assuming
               that it is undergoing pure rotational motion, as shown in Figure 6.11 (Stein 1995; Hartley
               1997b; Hartley, Hayman, de Agapito et al. 2000; de Agapito, Hayman, and Reid 2001; Kang
                                                                                    ◦
               and Weiss 1999; Shum and Szeliski 2000; Frahm and Koch 2003). When a full 360 mo-
               tion is used to perform this calibration, a very accurate estimate of the focal length f can be
               obtained, as the accuracy in this estimate is proportional to the total number of pixels in the
               resulting cylindrical panorama (Section 9.1.6)(Stein 1995; Shum and Szeliski 2000).
                                                                  ˜
                  To use this technique, we first compute the homographies H ij between all overlapping
               pairs of images, as explained in Equations (6.19–6.23). Then, we use the observation, first
               made in Equation (2.72) and explored in more detail in Section 9.1.3 (9.5), that each homog-
               raphy is related to the inter-camera rotation R ij through the (unknown) calibration matrices
               K i and K j ,
                                     ˜           −1  −1           −1
                                    H ij = K i R i R j  K  j  = K i R ij K  j  .    (6.52)

                  The simplest way to obtain the calibration is to use the simplified form of the calibra-
               tion matrix (2.59), where we assume that the pixels are square and the optical center lies at
               the center of the image, i.e., K k = diag(f k ,f k , 1). (We number the pixel coordinates ac-
               cordingly, i.e., place pixel (x, y)=(0, 0) at the center of the image.) We can then rewrite
               Equation (6.52)as

                                                 ⎡                −1     ⎤
                                                    h 00  h 01   f 0  h 02
                                     −1 ˜                         −1
                             R 10 ∼ K  H 10 K 0 ∼  ⎣             f       ⎦  ,       (6.53)
                                     1              h 10  h 11    0  h 12
                                                                f −1
                                                   f 1 h 20  f 1 h 21  f 1 h 22
                                                                 0
                                        ˜
               where h ij are the elements of H 10 .
   309   310   311   312   313   314   315   316   317   318   319