I want to transform perspective on an input image and output image from the new perspective. I have depth map and scale information. The picture roughly demonstrates what I want to do.enter image description here I have a single camera which is monocular I get an input image, I can estimate depth and scale information from that image using deep learning. The angle between the two perspectives is not large and two camera positions are on the same plane. Is there a way to get an image from the new perspective using the information I have ?

  • 1
    $\begingroup$ that'd be the normal "camera projection" mapping that you'll meet everywhere in computer graphics. $\endgroup$ – Marcus Müller Aug 9 '18 at 10:38
  • $\begingroup$ True, if I can calculate Homography matrices relating the both camera coordinates axes plane and the image plane this can be done. But the problem here is, the subject is a 3d object not a planar(2D) object. So using Homography introduces artifacts. That is why I have asked here to check if there is any other way to do this. $\endgroup$ – papabiceps Aug 9 '18 at 11:08
  • $\begingroup$ but: your object isn't really planar, is it? it's actually three-dimensional $\endgroup$ – Marcus Müller Aug 9 '18 at 11:11
  • $\begingroup$ Yeah the camera is capturing 3d scene (real world). $\endgroup$ – papabiceps Aug 9 '18 at 11:12
  • $\begingroup$ The bit that I am unclear about is what sort of "perspective shift" are you trying to achieve? Because once you have some sort of point cloud (or blob cloud), it should then be easy to rotate it to shift its perception slightly (?). But, that does not add / reveal any new information (?). $\endgroup$ – A_A Aug 9 '18 at 19:05

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.