0

I am trying to implement my own image stitcher for a more robust result. This it what I got so far.

enter image description here

The result of the panorama sticher that OpenCV provides is as follows.

enter image description here

Apart from the obvious blending issues, I am wondering how they distribute the warping to both images. It seems to me they project the images on some cylinder before the actual stitching. Is this part of the calculated homography or do they warp the images before the feature matching? I had a look a the highlevel of the stitching pipeline, the actual code as well as the landmark paper for the pipeline, but I couldn't figure out where exactly this warping happens and what kind of warping it is.

rex123
  • 371
  • 3
  • 16
  • 4
    It's not exactly "projected" to a cylinder; panoramas are assumed to be taken from a fixed location with the camera rotated for each photo, so you are *taking pictures of a cylinder* and it's projecting them instead to a plane. If it was projected to a cylinder, it would squeeze the top and bottom down more; but the opposite is true. [This](https://sites.google.com/site/ritpanoramaapp/project-stage-iii) might be an easier read than Brown and Lowe. Also there is a [very detailed OpenCV stitching sample](https://github.com/opencv/opencv/blob/master/samples/cpp/stitching_detailed.cpp). – alkasm Sep 26 '17 at 10:39
  • @AlexanderReynolds I was probably not very clear in my question. In the second picture above the images (each image by itself) looks like it was kind of warped onto a cylinder from the outside. Does this effect only come from the projection surface warping (step 7 of your first link) onto a cylinder? – rex123 Sep 26 '17 at 12:37
  • The OpenCV Stitching pipeline supports [multiple warping models](http://docs.opencv.org/3.3.0/d0/dfa/group__stitching__warp.html). – Adi Shavit Sep 26 '17 at 12:50
  • 1
    No, you were clear enough---but let's put that aside. Richard Szeliski is an author of one of the better computer vision textbooks, and a pioneer in image stitching/alignment/registration etc. He wrote a "tutorial" (it's like 80 pages) on image alignment but it's extremely awesome. You can check it out [here](https://people.cs.umass.edu/~elm/Teaching/Docs/Image%20Align%20Szel.pdf). Starting on Page 56 it discusses "composting" which is what you're asking about. – alkasm Sep 26 '17 at 13:09

0 Answers0