Converting the raw video from the Ricoh Theta SC requires knowing the lens parameters. Since Ricoh doesn't publish those, I set out to find them.
This is what a raw video frame from a Ricoh Theta SC looks like:
Two hemispheres - back and front - each circular with a radius of 480 pixels, aligned at the top of the frame. When converting this to something usable, two things must be corrected for:
Projection - resampling the image into a format that the panorama viewer can accept.
Alignment - correcting for imperfections in the dual-lens alignment.
There are also some other nice-to-haves, but these are not discussed here:
Nadir correction - The camera captures a bit of itself at the nadir point, where the overlap between the hemispheres is zero. This can be fixed by distorting the image slightly as you approach nadir.
Blending - In the presence of strong light sources the two cameras may get lens flares or other artifacts that makes a simple blend show strong seams (see below).
The projection parameters are quite straightforward.
|Projection||equidistant (R = f∙θ)|
|Field of View||202.15558 deg|
|Pixels / degree||4.7488|
|Image Center (front)||1440, 480|
|Image Center (back)||480, 480|
3. Hemisphere Alignment
The two lenses in the Theta are not exactly 180 degrees apart. The actual numbers will vary from camera to camera, but they can be found by taking the two hemispheres and aligning them in Hugin[a]. (I'm sure the information can be dug out of the Ricoh Maker Notes[b] as well, but Hugin is what I used.)
It suffers from a lens flare in the front hemisphere that probably should be smoothed out, but otherwise things line up.