I have a question regarding 3D reconstruction from the multiple view
images, which the text book did not explain very clearly.
Take 2-view stereo image for example. After the cameras are
calibrated, as long as we can find matched points from the 2 views, we
will be able to calculate the 3D depth of that point. However, not
very point in space has significant enough in both views. Say, for one
pair of images, we may find 300 matched points by features. But how
can we reconstruct the depth of the rest unmatched points (or
insignifcant points)?
Thanks a lot!