How did I create these panoramas?
- I shot 15 pictures in a circle, overlapping 1/3 of each shot
- Repeated another row angling camera up, then another row shooting down
- Shot one picture straight up (zenith), and one down (nadir)
- Assembled the ~50 images automatically with Microsoft's free ICE (image composite editor)
- Fixed nadir or stitching errors in the resulting jpeg with gimp (or photoshop)
- Uploaded it to 360cities.net
- Camera: Sony Cyber-Shot 25-250mm compact camera (14 MP)
- Set it to manual exposure and white balance, 5 or 10 MP, wide angle (25 mm)
- Use tripod with camera oriented vertically (~70 x ~50 degree FOV)
- Smallest aperture available (f 8) for maximum depth of field
- Adjust speed and if necessary ISO for good exposure
- Stitch the images with ICE. No intervention necessary or possible.
- Add the nadir text using gimp, erase stitching seams in blue skies if needed. With 10 mp individual shots, the final stitched image is too large to easily edit in gimp on my PC or mac. Photoshop may be able to handle them better.
- Any digital camera can be used. Even a 2 MP camera with multiple shots will produce reasonable detailed images.
- Good to have manual exposure --f stop/iso/speed. Otherwise variations in exposure in different parts of the scene will be hard to merge.
- Good to have tripod since if don't have fisheye or really wide lens, you'll need >30 shots
- Maybe good to have fisheye for 4- 10 shots or fancy 1 shot mirror lens (e.g. real estate virtual tour -- low resolution, no need for nadir/zenith.
- Few shots (as with a fisheye) are good if there are people moving around, since it's hard to keep them from being where seams occur. But fewer shots means lower resolution. The multigigapixel shots on 360cities.net are made using thousands of shots taken by telephoto lenses on cameras moved by computer.
- Stitching software -- I use Microsoft Research ICE: Image Composite Editor: simple, automatic and fast. If you need to mask out people or blend sky or nadir images etc., you can "export the images to photoshop" which gives you a stack of layers for manipulation. I have tried Hugin (freeware) which is semi automatic for assembly and blending, but takes a while to learn. I haven't been able to use it yet to assemble 4 full fisheye images taken with a Nikon FC-E8 adapter mounted on an old 2MP Canon Cybershot. Maybe a camera mounting problem, maybe my lack of expertise in using Hugin.
- Visualization software. 360cities.net requires high quality input, but will upload then automatically convert an equirectangular image created by ICE to two different tiled formats for display in Google Earth or via the krpano flash on their website. The tiling is necessary so that large images will load rapidly as small tiles. ICE can upload directly to Microsoft's photosynth.net, but photosynth requires MS Silverlight which has a low market penetration in comparison to the >99% claimed for flash. Can display panos with QuickTime (25% of computers?) or Java (25% of computers), but they don't using tiling and would be slow to download detailed images (my panos are 40 - 150MB (using 5MB - 10MB indivdual shots).