Curved Document Imaging with Eye Scanner Toshiyuki AMANO, Tsutomu ABE, Osamu NISHIKAWA, Tetsuo IYODA, and Yukio SATO 1. Shape From Shading SFS [1] [2] 3 2 Department of Electrical and Computer Engineering, Faculty of Engineering, Nagoya Institute of Technology, Nagoya-shi, 466 8555 Japan Intelligent Devices Laboratory, Fuji Xerox Co., Ltd., 430 Sakai, Nakaimachi, Ashigarakami-gun, Kanagawa-ken, 259 0157 Japan [3] Cho [4] CCD [5] [6] SFS SFS Cho D II Vol. J86 D II No. 3 pp. 409 417 2003 3
2003/3 Vol. J86 D II No.3 2.3. 4. 5. 6. 2. 1 1 Fig. 1 An exterior view of eye scanner. CCD [7] 640 480 1 CCD PC USB PC 2 334 PC USB RS-232C PC 3 2.1 2 Fig. 2 A structure of the camera head.
3 Fig. 3 System diagram. 2.2 600 A3 170dpi 3. 3.1 4 xc-yc xd-yd 4 Fig. 4 The coordinate systems of eye scanner. X Y Z X Y Z X 0 Y 0 Z 0 2 3.2 Tsai[8] Tsai [7] 3.3 Tsai
2003/3 Vol. J86 D II No.3 x c -y c x d -y d hx d hy d h = t x r 11 r 12 r 21 r 22 t y p 1 p 2 1 m x 0 0 0 m y 0 0 0 1 x c y c 1 + ε x x c, y c ε y x c, y c (1) 1 2 ε x ε y x c y c 2 4 ε x x c, y c =kx 00 + kx ij x ci y c j i 4 i Σi =1 Σj =0 4 i ε y x c, y c =ky 00 + ky ij x ci y c j i Σi =1 Σj =0 2 2 4 ε x ε y 0 (1) 1 x d (x c,y c ) y d (x c,y c ) 2 ε x ε y 3 3.4 (ω, φ)=(0, 0) X 0 Y 0 Z 0 = R φ R ω X V x Y V y Z V z + 0 (1) (2) V x V y V z (3) R φ R ω X YV x, V y, V z (ω, φ)=(0, 0) 4. 4.1 1 1 1
6 Fig. 6 A curved surface stretching. 5 2 Z = a 00 + a 01 X + a 10 Y + a 02 X 2 + a 11 XY + a 20 Y 2 (4) 2Z α Z = b 0 + b 1 X + b 2 X 2, X = X + cy α = tan 1 c Z Y Z (5) Z = d 0 + d 1 Y, β = tan 1 1 d 1 (6) X β 4.2 l E I y = 5 Fig. 5 The surface shape model of books. Mg 24lEI 6l 3 4lx + x 2 x 2 (7) 1 (7) 5 Z = k 0 + 5 Σ n =1 k n X n g(x) (8) (8) 6 g(x) X=0 X=X(x c,y c ) x c,y c X X(x c,y c ) x c,y c Z Z(x c,y c ) g(x) X X X Y 3 (1)
2003/3 Vol. J86 D II No.3 5. 5.1 7 1024 768 4.2 4.2 8 (8) (8) 8 Z 7 X(x c,y c ) Y(x c,y c ) Z(x c,y c ) (1) 9 9 7 Fig. 7 Newspaper texture on the cylindrical object. 8 Z(x c, y c ) Fig. 8 Range-image of the Fig.7 (depth Z(x c, y c )). 9 10 4. 11 7 11 210 mm 160 mm 144 dpi A B 1 mm 195 mm 196 mm
9 Fig. 9 Rendering result after pose normalization. 12 Fig. 12 Texture on color booklet surface. 10 Fig. 10 A result of lost region interpolation of Fig. 9. 13 Fig. 13 A stretch result of the color booklet surface. 11 Fig. 11 A stretch result of the texture on the cylindrical surface. 5.2 12 A3 4 13
2003/3 Vol. J86 D II No.3 14 Fig. 14 A result of document-imaging by the divided capture technique. 5.3 A3 500 mm 20 deg. 10 deg. 6 14 14 "management" 0.033 deg. 0.3 mm Z θ Z sinθ 5 mm 1.7 mm 100 dpi 0.7 mm 6.
[1] 3(I) Shape from Shading (D-II), vol.j77-d-ii, no.6, pp.1059 1067, June 1994. [2] 3 (II) Shape from Shading, (D-II), vol.j78- D-II, no.2, pp.311 320, Feb. 1995. [3] 3 (D-II), vol.j83-d-ii, no.12, pp.2610 2621, Dec. 2000. [4] S.I. Cho, H. Saito, and S. Ozawa, Shape recovery of book surface using two shade images under perspective condition, Trans. Inst. Electr. Eng. Jpn. C, vol.117-c, no.10, pp. 1384 90, 1997. [5] PRMU97-77, 1997. [6] M. Kashimura, T. Nakajima, N. Onda, H. Saito, and S. Ozawa, Practical introduction of image processing technology to digital archiving of rare books, International Conf. on Signal Processing Application Technology, 1999. [7] (D-II), vol.j76-d-ii, no.8, pp.1528 1535, Aug. 1993 [8] R.Y. Tsai, Versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses, IEEE J. Robotics and Automation, RA-3, pp.323 344, Aug. 1987. [9] T. Amano, T. Abe, T. Iyoda, O. Nishikawa, Y. Sato, Development of the next-generation document reader EyeScanner, Proc. SPIE, vol. 4669, no. 4669 05, San Jose, 2002. [10] ITS2001-38, IE2001-177, 2002. 14 5 27 10 17 12 12 3 12 62 TFT / 3D 52 57 / 4 9 3D 54 IT 3D 3 1980 1983 19951986 1987 2001 3 IEEE