PetaPixel

UT Austin Launches Free Enlarging and Denoising Web App

rcmwebapp2

Movies and TV shows have a knack for making it seem as if you could take a horrible, low-resolution image and turn it into a high-res masterpiece — the term “enhance” has become almost comical. And for every mention of magical television enhancement, there’s mention of some special algorithm at work that makes it happen.

Well, the University of Texas at Austin’s RCM Tools web app isn’t quite up to cable drama standards, but it’s their attempt to apply special algorithms to image enhancement and denoising, and it’s free for photographers to experiment with.

Ordinary enlargements contain JPEG blocking, checkerboarding, and ringing artifacts. The image on the right has been enlarged by the app with less artifacts.

Ordinary enlargements contain JPEG blocking, checkerboarding, and ringing artifacts. The image on the right has been enlarged by the app with less artifacts.

The web app is a free service provided by the Center for Perceptual Systems at the University of Texas at Austin that applies a computational method called recursive condition means (RCM) to perform certain image processing tasks.

The research is based around measurements of certain statistical properties of a large subset of “natural” images, and then uses those measurements to develop ways to better enlarge, denoise, deblur, color filter array interpolate, and compress images.

Enlarge photos with less noise by denoising with the app prior to enlarging.

Enlarge photos with less noise by denoising with the app prior to enlarging.

You can read all about the Recursive Conditional Means method by clicking here, but we’ll go ahead and warn you that it’s pretty dense stuff. For now, the site only allows you to play around with denoising and enlarging, but if you wanna try it out for yourself, you can do so here.

Image Processing with Natural Scene Statistics [CPS at UT Austin via Reddit]


 
 
  • sabrunto

    But photoshop already resamples you images when enlarging them.

  • http://tr.im/bomath Barbu

    Oh well… They made yet another denoising proggie which *doesn’t* work.
    Panasonic, in some of the most dreadful cameras when it comes to noise processing, would do ten times better. And yes, unless you consider DOWNsampling and then resizing back as “denoise”, you shouldn´t even try that site. Waste of time…
    I know I might be a bit (only a bit) too harsh on this academic endeavour, but there is absolutely no proof that their software is able at least to do better (noisewise) than the blur function in Photoshop.

  • guest

    But it’s got electrolytes!

  • wickerprints

    You have no idea what you are talking about. I actually read the paper and have the educational background to understand it. The approach is not without its flaws, but saying that “Panasonic…would do ten times better” is plain ignorance on your part.