Learning by Doing II

So I had to figure out how these mistakes happened. In order to understand that I had to go back to the Snow Globe model.















Looking at this picture I tried to figure out why the perspective corrections were not achieved by the software I used, to create the virtual 3D environment. (I need to mention here I am using this software called Pano2VR to twist and turn the panoramic image created in Photoshop in the virtual space).
So what did I miss?
What will I have to do to photograph the dome entirely from inside and then stitch photographs properly and gain a panorama that would play perfectly in QTVR or in the flash file?
Looking at the picture it occurred to me in a flash that I was not considering something very important while stitching the photograph. I totally omitted the distortion error created by the curvature of the dome. Its a complex 3D curvature. So the immediate realization that occurs is that there is another distortion factor to be taken care of and that is the barrel distortion created by the camera lens while shooting in extreme wide angle.
For the time being, I was at loss of wit, and tried to correct these distortions randomly through applying distortion and lense correction filters in photoshop. Obviously those didn't help, but there was a fair amount of learning on how the image distortions take place in the virtual 3D space. Here are a few examples:

(Move the mouse pointer inside the image window to move through the image, press shift to zoom, press ctrl to zoom out)

Experiment 1







For a full window view, go to
http://diplopanarama.99k.org/full%20pan%20curved/full%20pan%20curved.html

Experiment 2





For a full window view, go to
http://diplopanarama.99k.org/full%20pan%20sphere/full%20pan%20sphere.html

Experiment 2 distortion is done for a 3D projection option known as 'little sphere'. Here, the whole scene is fitted into a virtual 3D Sphere. The projection told me that the way I tried to fit the scene is wrong, the ground should have been in the middle, instead of the edge.

Experiment 3





For a full window view, go to
http://diplopanarama.99k.org/full%20pan/full%20pan.html

Curvature of the dome, and the distortion from the lens, these two factors when act in combination needs long sets of extremely complicated geometric algorithms to be rectified. Thinking of algorithms it immediately occurred to me that, as these long chain of calculations are beyond normal human brain to perform, so there will have to be some specialized software in order to perform this task. Thus began my quest for the software.