360-degree Boston Panoramas

In 2001 I collected a large number of "light probe" samples, in which a light probe (a mirrored sphere) is placed in an interesting environment (from a lighting standpoint) and a camera takes a picture of the probe. By taking pictures at multiple stops/shutter speeds, enough information is gathered to produce "high dynamic range" images which can be used to convcingly light synthetic objects placed in the scene.

However, despite collecting the raw data, I've never actually processed all my light probe images. This is because, with the cheap equipment and extreme zoom I was using, my light probe samples are not registered with each other, and I keep hoping somebody will write a high-dynamic-range conversion program that automatically registers the images. It doesn't seem to have happened yet, although at least somebody has written a paper about it.

So I decided I'm just going to use them for synthetic lighting without bothering to convert them to high-dynamic-range images. Instead, a dark image can be used for diffuse lighting, a mid-range image for specular lighting, and a bright image for chrome/mirrored surfaces. I do, however, need to process the light probe images to extract the actual data from the raw images.

As an interim step, I wrote a program to convert from the light probe image to a "latitude-longitude" panorama. The image wraps around horizontally; at the top and the bottom they stretch horribly, corresponding to the poles of the globe. Still, it's much easier to see the features in this form that in the original form--e.g. the example at the top of the page, the overpass above is nearly impossible to decipher in the light probe form--so I thought I would put up "panoramas" of all the light probe samples I took on my first day collecting them. I've only bothered to put up one of each of the multiple stops, that is, only one from each location.

Myself and a tripod are visible halfway between the center and the right edge of the image. Halfway between the center and the left edge is the "vortex of no data", the region that the mirror ball doesn't sample very well. Extending below that region you can usually see the rod from which the light probe was suspended.

The image covers 360-degrees around horizontally, and extends to straight up at the top of the image and straight down at the bottom.


It was overcast this day.

Morning, an intersection

Between two apartment buildings and under a tree

Approaching and within an underpass

A field

Stairs in an industrial area leading up to a highway

A manual car wash

On a bridge

Near the bridge, a tiny building

A small landscaped area between several buildings (the first two are the same, but a truck drives by)

A subway station that was vaguely near the last few areas... rather orange, was this an Orange Line?

Transferring at Lechmere, a Green Line station

Kendall, a subway station on the Red Line