FAQ

What are recent developments and improvements in hardware for scanners and digital cameras?

For this FAQ, we asked Kodak's Don Williams to look back over the past five years and identify significant new developments as well as important incremental changes in scanner and digital camera design, with special attention to the needs of libraries and archives. Don Williams is a senior research engineer in the Image Science Division of Eastman Kodak Co. He has written extensively about digital image capture specifications and imaging performance metrics and is a regular participant on digital imaging standards committees.

Recent developments and improvements in hardware for scanners and digital cameras

When asked to contribute an update on advances in digital image capture technologies to this forum, I hesitated momentarily, gauged my instincts, and accepted. After all, in the past five years all sorts of new image capture devices have been introduced from which I could draw. What could be easier? Then the penny dropped.

I realized, somewhat humbly, that there actually have been few fundamentally new approaches applied to digital image capture, especially for museum and library community level tasks. Rather then chasing promising but unproven new scanning technologies, most efforts have focused on perfecting existing ones. The good news is ... this is not bad news. Freed from the onerous learning curve of technology adolescence, manufacturers have concentrated on multiple incremental improvements that come with maturity. For the user, the impact is nothing but positive. Imaging performance, cost, and speed (think workflow) have dramatically improved. This benefits not only research organizations but, notably, resource poor local/regional sites with their own conversion tasks.

But, to suggest that nothing is new is remiss. Certainly some exciting technologies have been introduced and are implemented in a few products. These and the cited maturity improvements are briefly discussed below. Being a scanner gearhead at heart, I have chosen to organize these according to four scanner subsets. They are 1) document handling, 2) illumination, 3) sensors/detection, and 4) data processing. Some items may be scan mode (e.g., transmissive vs. reflective) or hardware specific (e.g., flatbeds vs. cameras) and will be emphasized as such.

1. Document handling—Two words immediately come to mind, "Book Cradles." This class of camera hardware has improved from the yawning, static, manual contraptions of several years ago to the robotically articulated page-turning wonders of today. Like any new technology there is likely to be an optimization period for these devices, but the forecast is good. See, for example, Conservation by Design's Preservation Book Cradle and 4DigitalBooks' ™ automatic digitizing system.
   Less seductive but equally pragmatic is the trend from cameras with horizontally constrained document placement to those with a vertical document mounting option. Gravity and conservator concerns have dictated this change. Although appearing to be a trivial modification to existing camera design, doing so while maintaining resilience, portability, and utility of the supporting structures can be a challenge, especially for very large documents. Nevertheless, many designers have achieved this adaptability with minimal compromise, some elegantly so.
2. Illumination—Though hardly noticed, illumination systems of flatbed scanners, both reflective and transmissive, have improved considerably. Largely, this is attributable to improvements in cold cathode fluorescent illumination sources used in these scanners. Their low cost, rapid warmup, stability, and improved color quality have made them nearly a universal choice for illumination sources in this class of scanners. Improvements in illumination optics and increased bit depth for these scanners have also provided dramatic uniformity performance.
   Several years ago it was advisable to avoid the platen margins on these scanners because of the uncompensated illumination falloff. Today, low cost scanners can be found where literally all of the platen area is effectively illuminated within 5.0 % uniformity. Epson flatbed scanners are particularly good, but any scanner can be tested simply by scanning a known uniform flat field document, like a Munsell paper sheet.
   From a conservator perspective, it is encouraging to see that some camera system manufacturers (e.g., Lumiere Technology) are proactive in designing ultraviolet and infrared friendly light sources for especially sensitive documents. The fading and heat characteristics of these portions of the radiation spectrum are a very real concern from a conservation perspective, particularly for high quality scans of long duration.
3. Sensors/Detectors—Despite the hype on the benefits (lower cost, higher level of feature integration) of CMOS (Complementary Metal Oxide Semiconductor) sensors several years ago, they continue to have inferior imaging performance (higher noise, lower dynamic range) than their CCD (Charge Coupled Device) counterparts. To my knowledge they are used exclusively as area array camera sensors and not as scanning linear arrays. This makes them perfectly suitable for many consumer or prosumer (i.e. professional consumer) camera applications but risky for demanding conversion projects. For this reason, CCDs continue to be used as the imagers of choice for conversion grade scanning applications. Several important changes to the sensor "imager package" are noteworthy. They can apply to either CMOS or CCD type imagers and are:
   
   1. Pigmented color filters—For color scanners where the color filters are coated onto the sensor, some manufacturers are beginning to use pigmented rather than dye based filters. The reason for doing so is the same as for using pigmented dyes in inkjet print applications—stability.
      
   2. Depth-wise color detection—This is a new color detection technology for digital cameras developed by Fovean. Its claim to fame is that it can capture a fully pixel-populated RGB digital image using a single area array detector in a single frame. Most of today’s studio cameras use scanning linear arrays (slow), color filter wheels with area arrays (requires multiple frames), or sparsely populated RGB color filter arrays (requires de-mosaic interpolation). Fovean has accomplished this by taking advantage of the well-known fact that different colored light penetrates to different depths within the detector bulk. Red light penetrates the furthest, green light less so, and blue light even less. By reading out the charge associated with different depths within the sensor one can in fact create a color image without the explicit use of color filters. This is not an easy task though and may require aggressive data processing to achieve the demanding image performance levels of imaging for the cultural heritage community. Currently cameras employing this technology cater to the prosumer market.
      
   3. Smaller pixel sizes—The individual sensors associated with a single image pixel have become progressively smaller over the years. Indeed, this has allowed prosumer/consumer digital cameras to increase their total pixel count without significantly changing overall detector size. Today, typical sizes may range between 3-5 microns per pixel compared to 7-11 microns of the past. These smaller sizes are not without their imaging performance tradeoffs. To achieve the same signal levels per pixel, about four times the illumination level is required (can you hear the paper conservators gasp?). Without these increased levels, a greater reliance is placed on subsequent image processing to deliver the image. Depending on the processing aggressiveness, this almost always increases image noise levels, which lead to lower image quality.
      
   4. Support Electronics—Perhaps the most impressive changes have come in terms of reducing the size of the camera/scanner’s support electronics. This is where the analog-to-digital conversion as well as much of the data processing (see next section) occurs. What used to be the size of a deck of cards has now been reduced, via CMOS integration, to that of a nickel.

4. Data Processing—Rather than cumbersomely performing image processing functions offline, there is a trend to integrate common scanner related functions such as OCR (Optical Character Recognition) and distortion correction within the support electronics. One of these functions, licensed from Applied Science Fiction (ASF) as Digital ICE™, is automatic scratch and blemish removal. It was first introduced for film scanners (Nikon) and more recently into reflection scanners (Microtek). Truly a technology change, Digital ICE™, relies on the scattering of infrared light by scratch and blemish artifacts in film and photographic paper. An infrared scan in addition to RGB color scans are made of the sample. The infrared scan is used to identify where the scratches are located. This information is then used to mask the blemishes through image processing in the other three color records. It works quite well for minor defects in color negative and incorporated color slide films (e.g., Ektachrome). Unfortunately, this technology has been known to behave erratically on film media common to the library and museum communities. For instance, mixed results occur for non-incorporated coupler films (e.g., Kodachrome) and it will fail completely on all black and white silver halide films.

Finally, a few words on multi-spectral or hyper-spectral image capture for artwork. In concept, performing these types of captures has always been easy. Through multiply-filtered frames and suitably designed light sources, a number of demonstration projects of this nature have been documented. (For some examples, see RLG Diginews, October 15, 1999.) But let’s face it, these projects have not been the epitome of productive workflows. They have, however, supplied critical examples of ways to improve the process and what shortcuts can or cannot be taken. Over the next several years I predict that large gains in productivity, economy, and quality will be made in this area of digital image capture. Some university and commercial partnerships are exercising new models for multi-spectral capture and it will be exciting to see the future levels of improvement.

Publishing Information

RLG DigiNews (ISSN 1093-5371) is a Web-based newsletter conceived by the RLG preservation community and developed to serve a broad readership around the world. It is produced by staff in the Department of Research, Cornell University Library, in consultation with RLG and is published six times a year at www.rlg.org.

Materials in RLG DigiNews are subject to copyright and other proprietary rights. Permission is hereby given to use material found here for research purposes or private study. When citing RLG DigiNews, include the article title and author referenced plus "RLG DigiNews." Any uses other than for research or private study require written permission from RLG and/or the author of the article. To receive this, and prior to using RLG DigiNews contents in any presentations or materials you share with others, please contact Jennifer Hartzell , RLG Corporate Communications.

Please send comments and questions about this or other issues to the RLG DigiNews editors.

Co-Editors: Anne R. Kenney and Nancy Y. McGovern; Associate Editor: Robin Dale (RLG); Technical Researcher: Richard Entlich; Contributor: Erica Olsen; Copy Editor: Martha Crowe; Production Coordinator: Carla DeMello; Assistant: Valerie Jacoski.

All links in this issue were confirmed accurate as of October 15, 2003.