0

I want to estimate the depth map between left and right images from "http://perso.lcpc.fr/tarel.jean-philippe/syntim/paires/GrRub.html". I understand that I must first calculate depth from disparity using formula Z = B * F/d The data set unfortunately does not provide Baseline distance B. Could you suggest how I can calculate this(if possible) or how I could calculate depth map from given data alone?

Thank you for your help. As I am new to stackoverflow and computer vision, do let me know if I should provide more details.

Priyamvadha Krishnakumar
  • 23
  • 6

1 Answers1

1

If you have the extrinsic parameters, rotational matrices R and translation vector t, there are two cases

a) (most probable) one of your camera (the main camera) is the centre of your coordinate system: the R1 matrix is the identity matrix and the related t1 is equal to [0,0,0]. In this case you could think the baseline B as the euclidean norm of the translation vector t2 of the other camera

b) in case none of your camera is the centre of your coordinate system, at least you should have calibrated your cameras with respect to the same reference system. The baseline B is the euclidean norm of the difference vector (t1 - t2)

(I was not able to open the left/right camera links, so I could not verify)

marcoresk
  • 1,461
  • 1
  • 12
  • 27
  • Intrinsic parameters(left camera): Image center: u0 = 379.775940, v0 = 284.423492 Scale factor: au = 1943.353760, av = 1885.454834 Image size: dimx = 768, dimy = 576 Focale: 1.0 Extrinsic parameters: Rotation: {{0.553452, -0.023944, 0.832537}, {0.007881, 0.999692, 0.023512}, {-0.832844, -0.006451, 0.553470}} Translation: {-1480.081421, 165.970978, -1214.238159} For right camera , all values remain same except for Translation: {-1216.573975, 170.300339, -1288.829102} – Priyamvadha Krishnakumar Sep 27 '16 at 17:33
  • I think I would be using the second option, since both are calibrated with respect to same reference system – Priyamvadha Krishnakumar Sep 27 '16 at 17:39