In the category of "eye-catching titles," the April 2010 issue of IEEE Transactions on Image Processing has an article entitled "Is Denoising Dead?" According to the abstract, the article asks "whether there is a theoretical limit to denoising performance and, more importantly, are we there yet?" Apparently the answer is no—not yet.
I think my optical stats professor said that the lower limit, or the best you could do, was defined by the Cramer-Rao Bound. It’s a bit tough to wade through, but there’s a discussion of it on Wikipedia. I think it basically says that no matter what you do, you can’t do better than this. Now I have no idea how close modern denoising algorithms are to that bound. Apparently that author thinks we can move closer to it.
Steve Eddins is a software development manager in the MATLAB and image processing areas at MathWorks. Steve coauthored Digital Image Processing Using MATLAB. He writes here about image processing concepts, algorithm implementations, and MATLAB.
I think my optical stats professor said that the lower limit, or the best you could do, was defined by the Cramer-Rao Bound. It’s a bit tough to wade through, but there’s a discussion of it on Wikipedia. I think it basically says that no matter what you do, you can’t do better than this. Now I have no idea how close modern denoising algorithms are to that bound. Apparently that author thinks we can move closer to it.
Mark—Thanks very much for providing the link.