Photographer Richard Sexton visited Better Light’s studio in San Carlos this past June to learn more about our large-format digital scanning backs. He also brought along his Canon EOS 1DS MK2 16-megapixel 35mm DSLR, and some 4x5 B&W sheet film, so he could make a comparison among images captured by familiar 4x5 sheet film, a popular DSLR, and a scanning back. Mr. Sexton details the photographic setup in a thorough article that can be found online at the www.luminous-landscape.com web site (full page address at end of paragraph). In the article, he describes how he uses two stages of additional interpolation and sharpening to increase the file size from the 16-megapixel Canon camera to approximately the same size as the uninterpolated file from the 144-megapixel Better Light scanning back. Several sets of “Detail” images are displayed, and the author states “I’ll let the results speak for themselves, i.e., I suggest you look at the pictures first and make some preliminary decisions for yourself before reading my comments.” He later concludes “The Better Light scan clearly has the most detail, though in many areas of the image it’s a subtle difference. Also, the color accuracy of the Better Light scan is the best overall.”  (full article: http://www.luminous-landscape.com/reviews/cameras/better-light.shtml)

Unfortunately, what isn’t stated anywhere is that the image “Details” displayed are not from the full resolution files described in the article. These “Detail” images have all been REDUCED IN RESOLUTION by amounts varying from 50% to 33% of their original size (one-quarter to one-ninth of their original data). Most readers will understand that if we interpolate an image up to several times its initial size, we haven’t added any real information, and if we subsequently reduce it by several times, we are pretty much back where we started. However, if we reduce a large uninterpolated image by several times, three-quarters or more of its original image information is discarded. This puts the scanning back at a significant disadvantage by eliminating its most desirable feature — much higher resolution. After examining the reduced-resolution “Detail” images in Mr. Sexton’s article, some readers might conclude that a Canon DSLR is capable of producing images nearly as good as those from a Better Light scanning back, regardless of what the article says.

Below are similar “Detail” sections from the original 48-bit RGB images described by Mr. Sexton in his article, reproduced at 100% of the finished image size. Because these “Detail” images are much larger than the reduced-resolution versions in the article, only a portion of each “Detail” is displayed. These sections were each cropped from the original images, converted to 24-bit RGB, and saved as high-quality JPEG files. Interested readers can save these image sections, and open them in an image editor where they can be magnified for even closer inspection. Subtle image quality differences often become more apparent at 200% magnification, although JPEG compression adds minor artifacts that also show up more when magnified. Some readers might question the validity of such close examination of an image file, but anyone interested in obtaining the highest-quality images soon learns to appreciate this level of scrutiny. The Canon DSLR image is very well-behaved, but it still only contains a fraction of the information present in the Better Light image — one-ninth, to be exact. Just as it was with film, the improvement in image quality made possible by a large-format scanning back may not be needed for some applications, but photographers pursuing the absolute best should pay close attention.

>> NEWLY ADDED – equivalent-resolution scans of Fuji Acros B&W film, as described in the article. Mr. Sexton mailed these to me from New Orleans just before Hurricane Katrina struck, and they didn’t arrive in time for the original article posting. I was surprised and pleased when his CD-ROM showed up a few weeks later, and we have now added the corresponding “Detail” sections from the 4x5 B&W film to this article, along with a few comments.

Slight differences in size between the two cameras’ image sections are due primarily to differences in perspective — the view camera used swing, tilt, and ample rise to keep the subject square, but the DSLR captured the image without these camera movements. Parts of the subject closest to the DSLR lens are bigger than parts that are farther away, so it’s impossible to exactly scale one entire image to the other. It’s not clear that the view camera lens isn’t introducing some distortion of its own, either — in retrospect, we should have operated both cameras “straight on” for a closer scale match. This would also let the scanning back use the best (center) part of the lens, instead of being well off-axis. The images immediately below have each been scaled down by the same factor from the original 48-bit files described in the article, and the same size rectangle has been drawn on each image to help show differences in scale and perspective. (The B&W film image used the same 4x5 camera and lens, and was scanned and cropped to match the Better Light image size and resolution, so it also has the same perspective and scale.)

Above are MODIFIED sections from the deepest shadows of the original full-resolution files described in the article — these sections have been brightened up by aggressive Levels adjustments to more clearly reveal noise and other potential image quality problems.  (Note that JPEG-compressed images don’t take kindly to aggressive Levels adjustments — all you see is JPEG noise — so we adjust the original 48-bit files, convert them to 24-bit RGB, and then save the adjusted sections as high-quality JPEGs.)  The difference is pretty obvious here – the Canon image maintains shadow tone separation fairly well, but it has significantly more noise than the Better Light image, even though the DSLR was operating at about half the ISO sensitivity of the scanning back — ISO 160 for the Canon vs. ISO 300 for the Better Light.  The 1DS MK2 may have good low-light sensitivity, but this is not the same as having a wide dynamic range, where the large pixels of the scanning back provide a decided advantage. 
 
Except for a few pesky dust spots, the modified B&W film image does pretty well in the darker tones, but graininess becomes apparent higher up the tonal scale.