Technical information about the panorama

Creative decisions first, of course. The gallery space lends itself to a cylindrical panorama (the floor and ceiling aren't that interesting!). Apple's QuickTime VR is the obvious choice, since it is by far the most prevalent format—meaning many users won't have to download a plugin—but only supports cylindrical panoramas. Although the gallery is rectangular, a large freestanding "island" in the exhibit precluded a single-node approach. Four nodes cover the entire space with no obstructed views.

The images were captured by a Nikon CoolPix 950 digital camera equipped with the Nikon WC-E24 wide angle converter. The camera was mounted on a Kaidan Kiwi 900/950 panorama head and Kaidan QuickTilt Leveler, supported by a Gitzo 1227 tripod. The CoolPix 950 was set for 640x480 resolution, and the Autoexposure Lock (which locks both exposure and white balance) was used. The white balance was sampled off the back (white side) of a Kodak gray card in the ambient light of the gallery, and the gray card was included in several "extra" exposures to provide a guide for correcting color cast later on.

Images from the CoolPix 950 with the wide converter exhibit pretty severe barrel distortion, made all the more obvious in this environment by the short distances inside the gallery. Uncorrected, these images would be difficult to stitch accurately. Prior to stitching the barrel distortion was corrected using Helmut Dersch's excellent Panorama Tools. Prof. Dersch has a useful example of correcting barrel distortion on his Web site, although his example uses images in landscape orientation. For the portrait images used in this panorama, the values of a=0 b=0 c=-0.086 d=1.086 worked best; your mileage may vary. Plug: Prof. Dersch makes his tools freely available, which is an incredibly generous gesture and a real service to the VR community. Thanks Prof. Dersch!

The images were stitched with Apple's QuickTime VR Authoring Studio (QTVRAS). After stitching, the images were brought into Adobe Photoshop 5.5 for color correction and sharpening. Back in QTVRAS, the images were converted into panorama format, and the four nodes were joined into a multi-node panorama (a "scene" in QTVRAS lingo). Hotspots for node-to-node navigation and as links to descriptive pages elsewhere in the site were added at this stage, also in QTVRAS. Here is a screen shot of the Scene Maker setup; the map is from the same Illustrator file described below.

Low-resolution previews (generated from the node images themselves), node labels, and user hints were added in conVRter, from VRTools.

The gallery map was drawn to scale in Adobe Illustrator from the exhibit designer's plans, using only Web-safe colors. The Illustrator file was rasterized in Photoshop at two different sizes, one for the QTVRAS Scene Maker background (I know, it was a frill) and one for the panorama node map. The node map image was rasterized at exactly 320 pixels wide, the same width that was chosen for the panorama display setting; this way the panorama window and its map stack perfectly. The text labels were added to the node map, and each map image was compressed to a GIF, all in Photoshop.

The map image was added to the multi-node panorama using mapsaVR from Squamish Media Group.

The finished multi-node panorama with map was optimized for Web delivery using deliVRator, also from VRTools. This step reorders the various media tracks, ensuring that the low-res previews stream first (among other things).

Questions or comments are welcome.