Help on reverse panorama transformation

[Noveguy]

Hi all, 

I would like to ask about:

1. How Hugin fill the overlapping area between 2 single images of panorama ?

2. Is there a way to get the most correct reverse mapping of one pixel from panorama to single image ?

I am doing experiment of reverse transformation from panorama to single image.

My rig has 6 fish-eye camera to capture 6 single images, then I used Hugin to stitch 6 single images into 1 panorama.

I did try reverse transformation from panorama to single image using pano_trafo.exe

The result was like below:

PanoMapping::reverseMapping: 0 17.8527 4761.12

PanoMapping::reverseMapping: 1 -537.378 -1672.43

PanoMapping::reverseMapping: 2 4063.91 -1364.64

PanoMapping::reverseMapping: 3 3195.1 298.536

PanoMapping::reverseMapping: 4 964.359 1025.45

PanoMapping::reverseMapping: 5 657.126 3220.11

I filtered the above results to get 2 valid solutions:

#3 3195.1 298.536 (valid but wrong)

#4 964.359 1025.45 (valid and correct)

So I had total 6 results (0-5) but only #3 & #4 were valid (since my picture size is 4056 x 3040, the returned pixel coordinate could not negative or out of picture's size)

I did visual check and I confirmed that the #4 is the same pixel that I chose on panorama (#3 & #4 is corresponding to image #3 & #4 of my 6 single images)

I repeated this experiment on other pixels and my conclusion is that:

At the overlapping area between 2 single image, Hugin seems to choose one image as "dominate" one and use that "dominate" one to fill that overlapping area because there is always one very correct (both on coordinate & visual check) returned from pano_trafo.exe. If so, how can I get the information of that "dominate" one so that I could choose the correct mapping ?

Thank you for reading,

[T. Modes]

Hi,

Noveguy schrieb am Sonntag, 16. Januar 2022 um 08:06:51 UTC+1:

At the overlapping area between 2 single image, Hugin seems to choose one image as "dominate" one and use that "dominate" one to fill that overlapping area because there is always one very correct (both on coordinate & visual check) returned from pano_trafo.exe. If so, how can I get the information of that "dominate" one so that I could choose the correct mapping ?

in a first step all input images are remapped into the panorama space. Then these remapped images are blended with enblend. Enblend blends the images depending on the image content and the size of the overlapping region. Even pixel in the neighbourhood of the overlapping region can be affected by the used Burt-Adelson multi-resolution spline algorithm. So there is no "dominating" image. In the overlap the result is always influenced by all overlapping images.

From a geometrical point of view all remappings are equally correct. There is not "correct" or "wrong" remapping.