The above water shots were the standard pano shooting. Underwater - We have a site we're building just for U/W panos - http://digitaldivemaster.com/
The goal there is to have every popular shipwreck in Lake Michigan shot - it will take time.
I made the Buccaneer site to help promote the new wreck, and it will eventually be a showcase for my business. Once I learn all the KRpano code.
Take a look at....http://www.digitaldivemaster.com
for photos and details of the pano rig . We use a 10mm fisheye, but are looking at an 8mm. The problem is that underwater housings need to be put down out of the water. Traditionally they rest on the lens port - so the lens port (technically a dome port) has protective ridges that resemble a butterfly shade. This is heavy duty, either plastic or aluminum, and you really don't want to trim it. The dome port will get trashed after the first dive. That actually happened to me years ago - I handed my new, $3000 housing with the fancy $1800 dome port up to the divemaster so I could exit the water. He set it on the dome port. It had no protective shields. I was able to sand, buff and polish out the scratches, but it's not something I would want to do again! Plus the 8mm dome port costs about 3x the price of the lens....
So 8 shots, plus an up and down work best. The down shots are easy - since the diver hovers, there are no feet or legs in the photo. The up shot is also simple. The underwater shooting is about the same as an above water shoot. We're experimenting with focus. Usually above the water I can set the focus to infinity, or around 30 feet. Underwater it's not working - I think the magnification factor of the water and the fact that the image translates from water / dome port / air / lens is effecting the fixed focus. Autofocus works - unless there's a large chunk of algae floating by - then it focuses on that! Since it's underwater, using the camera screen to preview the photo is pretty useless. The gap from the camera back to the diver's face is pretty big, and you're looking through the acrylic housing, water, the glass of your mask (plus any fog it has).... Bottom line - you don't know what you have until the images are opened on the computer. I have multiple shots on two different wrecks that are out of focus. If I disable the zoom in the viewer, they look ok. These will have to be reshot - but it's expensive. It costs $120 for the dive, plus air, travel, etc. So we have to invest about $200 to shoot the images. I know this does not sound like a lot, but it totals about $1000 a month!
As for stitching, etc. That's a pain. I tried PTgui and Autopano - neither can handle the images. PTgui found 2 control points, Autopano found 6. That's across 10 images! So I manually assign control points. I use Pano2VR to pull out the zenith and nadir shots, defish the up and down photos, then manipulate them into place in Photoshop. This involves opening each of the two shots Pano2VR extracted, pasting in the defished images, making everything fit, and then using Pano2VR to reinsert the images into the pano.
One thing we're going to try is to bring along a small bag of 1" stainless steel washers. These will be placed throughout the scene before shooting. Hopefully control points can then be automatically generated. If not, the washers will make it a lot easier for me to assign points manually. Some scenes still require a lot of touch-up work. The aft for example - that wooden section that's at an angle is chained to the deck, but bobs around. So it's in a different place in every photo - I have to manually remove the ghosts.
But the work is worth it. We're adding a spring detent pin to the pano head base. Once that's done I should be able to make a master template and apply it to the images - then I won't have to worry about control points at all.
Thanks for the supportive comments!