Bayer interpolation: Joining the dots
By
Mike Tomkins
(Monday, October 21, 2002 - 23:13 EDT)
We've been meaning to mention this for a little while now, but what with the constant flow of behind the scenes work are only just getting to it.
Mike Chaney of Digital Domain Inc., author of Qimage Pro and Profile Prism, has put together an excellent page comparing the results of Bayer interpolation with samples from Sigma's SD9 digital camera - a model which has provoked much excitement for using Foveon's X3 sensor technology that measures levels for all three colors at every pixel.
Mike has taken the Bayer interpolation he wrote as part of the RAW conversion software for Qimage Pro, and fed it a couple of uncompressed sample images from the SD9, courtesy of our friends at the Digital Photography Review website. The comparison isn't perfect (few are) - if anything it gives a slight edge to the Foveon chip since the Bayerised samples are created from the Foveon ones, and hence would carry through any flaws or processing (white balancing, contrast adjustment, sharpening, etc.) inherent in the Foveon images.
With that proviso though, we think the article goes a long way to helping readers to understand how Bayer works and why anti-aliasing filters are a must in most Bayer designs. It also suggests what differences to look for when comparing real-world Foveon and Bayer samples, and where Foveon's sensors may see advantages over the competition. Equally, it is an impressive demonstration of what can be done with Bayer interpolation, considering that literally two thirds of the image data is being thrown away and then guessed from the remainder!
For more, visit the DDI Software website to see Mike's article in its entirity.
|
