0

referring to the question about fundamental matrix,If I had a stereo pair (2 jpeg) and I want to apply Peter Kovesi's or Zisserman's function on order to obtain F, how can I retrieve P1 and P2 ? these two matrices are 3x4 matrices from the two images, but I don't know how they are related... is it right if I take a random 3x4 matrix from grayscale first image and the corrispondent 3x4 matrix in the second image obtained from the first one by using some matching technique such as correlation ? and if it is, do you think that a 3x4 matrix is not detailed enought?

Community
  • 1
  • 1
user2614596
  • 630
  • 2
  • 11
  • 30
  • Those 3x4 matrices are projection matrices, not F. F is typically used to retrieve E which is used to rectify a stereo pair or figure out the rotation and translation of the camera up to a scaling factor. You have to tell first what do you want to do, and why you use a stereo pair. – Vlad Mar 10 '14 at 14:17

3 Answers3

2

The Computer Vision System Toolbox includes a function estimateFundamentalMatrix that does what you need. Check out this example of how to estimate the fundamental matrix, and then use it for stereo rectification.

Dima
  • 38,860
  • 14
  • 75
  • 115
1

Well, most of your doubts would be clarified once you go through Camera Calibration and 3D Reconstruction.

Richie
  • 4,404
  • 2
  • 17
  • 22
0

You should not try to construct F from C0 and C1, you should rather use feature matching between the two images and compute F from the matched key points.

You can then set C0 to be eye and compute C1 from F

Rosa Gronchi
  • 1,828
  • 15
  • 25