I'm trying to understand 3D image reconstruction, and want to clarify how we can find a z image coordinate.

Is the z coordinate for image coordinates (x,y,z) the focal length since that's the distance to the focal plane? If so would using z=(fx+fy)/2 from K = cv.calibrateCamera() work (K~camera matrix)? Normally images are in normalized coordinates (x',y',1) where z=1. Do I multiply (x',y',1)*z (where z=f), or just use z=1?

Some additional questions if anybody can answer them:

From GPU Parallel Computing of Spherical Panorama Video Stitching, it states "x, y, z are coordinates of Cartesian coordinate system in image pixels", where spherical coordinates are calculated by theta=arccos(z/sqrt(x^2+y^2+z^2)), phi=arctan(sqrt(x^2+y^2)/z) and I assume I can just use regular pixel coordinates after centering them at (0,0). Is this correct?

Additionally from a Szeliski paper, Creating Full View Panoramic Image Mosaics and Environment Maps: "we map world coordinates p =(X; Y;Z) ...to 2D spherical screen coordinates (theta,phi) using: theta=arctan(x/z), phi=arctan(y/sqrt(x^2+z^2)". What does he mean by "world coordinates". Does this mean the image plane in world coordinates as nowhere in the paper does it call for real world object point positions?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.