ACQUISITION

Visual and 3D information present in the world surrounding us can be acquired in numerous ways. They include devices such as high-quality still cameras, video cameras, depth sensors (Kanade 1994, Michelangelo Project n.d.) multicamera arrangements (Narayana et al. 1998) and omni-directional sensors (Nalwa 1996, Nayar 1996). Different acquisition systems can be used for digitizing different objects. Rare books are digitized using a still camera set in a fixed position, whereas an object such as a statue is imaged from several different viewpoints. Issues that need to be resolved as part of acquisition are image resolution, cost, acquisition time, revisiting the acquisition process, the number of vantage views, synthesis of views from directions not included in the scanning and mixing of synthetic and real content.

The panel saw examples of acquisition of digital content at three sites: Keio University, the National Diet Library, and the National Museum of Ethnology. At Keio University, the emphasis is on digitizing rare books and manuscripts3/4Western, Japanese and Chinese. The Keio University Library has a large collection of rare books, one of its prized possessions being a Gutenberg Bible.

Keio University uses three different cameras for digitizing rare books, illustrations and photographs. A high-speed digital camera, an NTT-Olympus prototype, is used to digitize an entire book very quickly. It takes about five sec/page capture and display a single page. Figure 6.2 illustrates one of the camera setups at Keio University. Illustrations in the rare books are scanned using a high-quality camera that takes about two minutes/page. A third scanner, Kodak professional PCD scanner 4045, is also available for scanning photographs.

Fig. 6.2. Camera setup for scanning rare books.

The National Diet Library (NDL) has one of the largest collections of digitized rare books, documents and children's books in the world. The NDL out-sources digitization work to private companies and does not possess any acquisition devices.

The National Museum of Ethnology has an operational setup using three 3D scanners that can handle 100 objects per day in an "assembly line" fashion. These scanners can image objects that vary in sizes from 40 cm to 100 cm and can also produce measurements such as height, width and depth. Although dense imaging of 3D objects is possible (at 1/2-degree increments from all directions), typically five images from front, back, left, right and top are collected.

It is the panel's impression that the Japanese sites have shown operational capabilities in acquiring a wide range of artifacts, rare books and illustrations. There is a general appreciation for preserving rare books and other objects, and the Japanese even seem to be willing to subject their rare collections to repeated scanning to keep pace with improving acquisition technology. Keio University researchers cite the superior lighting system used for imaging the rare books. This "cold lighting" system assists in the preservation of the books rather than causing them to deteriorate.

One or two 2D images of pages from a rare book are often enough to create the necessary digital content. For 3D objects such as statues, many more images from different directions are required for virtual reality presentation. Although facilities for acquiring images at 1/2-degree increments from all directions are available in the National Museum of Ethnology, few images are collected in operation. One of the ways in which the paucity of data can be handled is by developing techniques for synthesizing new views from existing ones (Levoy and Hanrahan 1996). Such capabilities were not in evidence at the museum.

Another important issue in the acquisition process is cataloging. Different levels of capabilities were demonstrated. The National Ethnology Museum has the most streamlined cataloging scheme for 3D objects. On the other hand, the National Diet Library, although it has millions of pages imaged, does not yet have a cataloging mechanism in place.