Panoscan takes on a whirlwind assignment, improves upon its original product

by Hank Russell

November 2003

Back in 1997, Ted Chavalas was working for a research company in northern California. He tried to find a way to make a virtual reality movie. (This was before QuickTime VR ever came out.) From that, he designed Naviscan. "It was basically a pan-and-tilt head for a video camera that grabbed frames, and we eventually stitched all those frames together for a virtual reality movie," he explained.

In January the following year — in the middle of the dot-com boom — the company Panoscan (Van Nuys, California) was founded. Chavalas, the company’s president, was looking at linear line-scan sensors after noticing that realtors’ Web sites gave visitors virtual tours of the homes for sale.

"I was trying to find a way to make a digital camera that would make cylindrical or spherical images of a space," Chavalas said. "That’s how the camera started."That camera was the Mark I.

Photo: The MK-2 from Panoscan was used in shooting the interior of the space shuttle simulator and military and civilian aircraft. These panoramic images were later used in interactive training CD-ROMs for fire and rescue personnel.

The Mark I (MK-1) prototype was up and running in 1997 and hit the market by January 1999. It was developed in partnership with Phase One — a Danish company known for its 4x5 view scan backs for studio photography and design — which was the OEM in supplying the sensor board and electronics for the MK1. "Our camera is not really a camera; it’s a scanner," Chavalas emphasized. "It works very much like a flatbed scanner."

The MK-1 used a Kodak tri-linear sensor, which is 7,072 pixels tall and does true RGB color sampling. "There’s no Bayer pattern filters, there’s no color interpolation," Chavalas said, "so we can achieve much larger files and get better color and better quality."

It can capture a maximum image size of 7072 x 44,000 pixels and a maximum color of 14 bits per color channel, 42 bits per pixel. Its ISO ranges from 100 to 1,600 and has exposures speeds between eight and 125 lines per second. It has a vertical field of view of 45° to 180° and a horizontal field of view of 1° to 390°.

The lowest resolution image (884 x 2800) could be scanned in one minute. A low-resolution image (1768 x 5600) and a medium-resolution image (3536 x 11,200) can be scanned in two and five minutes, respectively. A high-resolution image at 7072 x 22,400 took 12 minutes with the MK-1. Images are saved in 24- or 48-bit TIFF format, and the MK-1 interfaces with either SCSI or IEEE-1394.

"Primarily, we sold a lot of our cameras for shooting automobile interiors for the automotive industry,"Chavalas said. "As the bandwidth has gone up, the quality of images has gone up. They don’t want these low-resolution images. These images are typically quite detailed because people want to be able to zoom in and see the detail of the car interior."

NASA used the MK-1 to produce panoramic images inside the space shuttle for a training CD-ROM entitled DDMS Space Shuttle Fire/ Crash/ Rescue procedures, which is distributed to personnel in the United States and eight foreign countries stationed at 20 emergency landing sites to learn and review emergency procedures. SIGMAinteractive Technologies, a high-tech aerospace and information solutions provider based in Huntsville, AL, produced the CD-ROM. The DDMS (Department of Defense Manned Space) Training Division is responsible for ensuring that DoD forces around the world are trained to support Space Shuttle contingencies.

On September 9, 2001 — two days before the terrorist attacks took place — Chavalas went to the Johnson Space Center to shoot inside the space shuttle simulator. "They (NASA) were very concerned about the safety issues," he explained. "About a month before the shoot, we started to have weekly conference calls with the safety review board and write up procedural information for them so they can analyze exactly what we were doing, what equipment we were bringing in, how much heat load would our lights produce inside the vehicle. It was definitely an interesting experience."

"The part of the problem is, when you’re shooitng in an environment like that and you’re shooting a spherical image, you need a lot of depth of field because there’s a wide range of distance," Chavalas continued. "Some things are very close to the lens, some things are very far from the lens, so you have to shoot it at f22 or f16 (stop). To do that, you need a lot of light, of course — more than is normally available inside there — so we had to augment the lighting."

Panoscan’s latest offering is the Mark II (MK-2). "Our second camera was designed completely from the ground up," Chavalas said. "The camera body, we designed; the electronics and the CCD, we OEMed from Better Light." Not only is the MK-2 four times faster than its predecessor, it is also four times more sensitive (it can go ISO 1600 and beyond — even at full resolution). It can scan more than 120 lines per second at full bandwidth or resolution and shoot at 585 megapixels. It uses Better Light Viewfinder 5 software and a Kodak KLI 6013 tri-linear sensor. Its triple pipeline architecture consists of three digital-to-analog converters and its processor module has a self-contained hard drive for storing images ripped into a .TIFF format.

For sharper images, Mamiya medium format lenses are used; it also accepts standard 62mm creative filters, which are located behind the lens. Exclusive anti-reflection coatings reduce CCD glare. Like the MK-1, the MK-2 can interface with SCSI or FireWire. A portable power source allows for six to eight hours of continuous use.

The MK-2 was designed based on customer feedback. "A lot of them (the requests) had to do with ease of use," Chavalas said. "Our first camera didn’t have any kind of shutter inside, so you had to dark current calibrate it before each scan by putting the lens cap on. In the new camera, we added an automatic shutter so you don’t have to do that anymore. It makes it much easier to use."

So far, the automotive, real estate and hotel/ hospitality industries, military and law enforcement have been using MK-2. In addition, the Golf VR Web site allows golfers to virtually tour any of the PGA golf courses, as well as each hole on that course. "Obviously, it’s for people who have $30,000 to spend on a system," Chavalas said. "It’s not a hobbyist’s camera by any means, but if you’re doing serious work, it’s a great tool for the job."

Chavalas said he is looking to go further in the resolution department. He wants to upgrade the tri-linear sensor to the KLI 8023, which is larger than the KLI 6013. But he is concerned that too much resolution may be just that — too much.

"It’s really where we’ve reached the point where it’s almost too much resolution," he said. "You reach a point where you outresolve the lenses. My personal belief is that these CCDs that have pixels smaller than 10 microns really, other than hitting the sweet spot in the center of the lens, don’t ever really achieve their full resolution. Consumer cameras that are five megapixels usually outresolve the lenses because the pixels are too small."