Tuesday, May 22, 2007
The Art Of The Up-Res
Getting beyond the limitations of your camera's native resolution is an art and a science
Camera Raw Vs. Bicubic
When digital cameras first started showing up in the marketplace, they were considerably smaller in capture size and not nearly as clean a capture. At that time, a method of up-resing became a common technique: Step Interpolation. The technique involved using multiple passes of Image Size in Photoshop to apply a 10% incremental increase in resolution per pass. This amazed many people, including myself. How in the heck could doing multiple passes be better than a single pass? I didn't know, so I passed along the technique to Photoshop engineer Chris Cox to figure out why it seemed to work. As any good pit bull would do, he sank his teeth into the problem and discovered a reason.
The term “bicubic” means that the interpolation method was looking on two axes (the “bi”) and then using a sub-sampling of four pixels (the “cubic“), more or less. So running multiple passes of Bicubic had the effect of increasing the sub-sampling and resulted in a superior up-res interpolation, particularly for digital captures from CFA cameras with Bayer array sensors. In addition to the up-res, there was a smoothing effect that helped digital captures.
At the same time Cox was working on improving the interpolation scheme for Photoshop CS, Thomas Knoll was independently working on an up-res scheme for Camera Raw. While not as extensive and complete as Photoshop's Image Size function, Camera Raw's interpolation was better than Photoshop's. Remember that Camera Raw first shipped during the Photoshop 7 time frame. Working separately on the same problem, both Cox and Knoll arrived at very similar solutions—an interpolation algorithm that's optimal for smaller images without a lot of very high–frequency film grain—essentially, digital captures. As a result, whether you up-res in Camera Raw or later in Photoshop CS or CS2, you get a scheme well–designed for the purpose of up–resing digital images. The importance of the smoother up-res is the ability to attack the image with more aggressive image-sharpening routines. Some up-res algorithms already have a component of sharpening built in. As a result, though Camera Raw and Photoshop's up-resing may appear smoother than other schemes, in fact, they're capable of a much higher level of sharpening without introducing artifacts.
The top image was up-resed in Photoshop CS2 using Bicubic Smoother to get from a 6-megapixel native resolution to a nominal 25-megapixel interpolated file size. The bottom image was up-resed in Camera Raw 3.1 using the 25-megapixel interpolation option. Both images are being displayed at a Photoshop screen zoom of 1200% to show the results of each method. They're essentially identical for all practical purposes.
Well, I told you that this whole thing was going to be problematic, remember? But I think you get the point. If you start with a technically correct image and do the proper image processing, you can get a good result up-resing even up to 400%.
Jeff Schewe is a professional photographer and all-around digital guru. Part of the Alpha testing team for Adobe Photoshop, he's a founding member of PixelGenius and the editor of PhotoshopNews.com.