The University of Texas at Austin
  • Shh! Your Photos Are Too Noisy

    By David A. Ochsner, College of Liberal Arts
    Published: May 29, 2013
    Before and after images that were fixed in Wilson Geisler's de-noising website.

    The images on the top exhibit “noise,” while those on the bottom have been corrected using Wilson Geisler’s photo-correcting website. [Photos courtesy of Wilson Geisler.]

    Don’t you hate it when your cell phone pictures are noisy? That is, when the lovely soft-light scene you sought to capture with a digital camera comes out looking muddy, fuzzy, grainy or pixelly?

    In digital images, that distracting fuzz is called “noise,” and it’s akin to the extra signals that cause the audible static noise when you listen to old AM radios. Digital cameras, unlike film cameras, are susceptible to a similar kind of electrical fluctuation, but the resulting noise here is visible.

    And while those imperfections can be annoying when you’re taking pictures at your child’s birthday party, too much noise in an image taken, say, by a satellite, can prevent important information from being seen and understood.

    Enter Wilson Geisler, who has developed a new tool to reduce noise in images. The free image-processing website, which allows users to upload up to 1,000 images daily, provides tools to “de-noise” images — such as removing imperfections resulting from low light conditions — and also allows users to enlarge images without losing picture quality.

    Geisler, director of the Center for Perceptual Systems (CPS) and a professor of psychology, says the algorithm makes corrections based on what it has “learned” from examining so many images.

    “We are taking a different approach and measuring statistical properties, understanding the structure of what is noise and what is not,” he explains. “This tool analyzes images more accurately because it is based on measuring the structure of natural images.”

    While other photo enhancement tools work on an image-by-image basis, Geisler’s approach — called image processing with natural scene statistics — is based on the analysis of thousands of images. Geisler and his team, including Jeff Perry in the CPS, originally set out to understand how animal and human visual systems work.

    “There is structure in the natural world. You learn that structure,” Geisler says. “As an organism, if you have learned this structure you can perform better than an animal that hasn’t learned it. The same applies if you want to build a better camera.”

    In addition to improving our everyday photographs, the technology will have applications in enlarging and enhancing everything from archival photos to satellite images. Geisler says those working with particularly large images will benefit from the speed of his service.

    “Compared with other photo-enhancement algorithms, we believe this is the best in the world for reducing noise and enlarging images, and also about 50 times faster,” he says.

    The site is easy to use. After agreeing to terms of service, users can immediately begin uploading photos and enhancing them in either “standard” or “expert” modes.

    Quick, get those birthday, beach and graduation night photos uploaded, enlarged and clarified!

    Before and after images that were enhanced using Wilson Geisler's photo site.

    More detail is visible in the photo on the right, which was enhanced and “de-noised” by Wilson Geisler’s photo site.

    Find more technical information on how the tools were developed.

    • Quote 2
      Brian said on June 6, 2013 at 6:52 a.m.
      Excellent work! I'm looking forward to using the website!
    • Quote 2
      Marsha B said on June 3, 2013 at 12:16 p.m.
      This is revolutionary concept and good timing as i just returned from Europe with hundreds of photos from iphone and compact digital camera. I am sharing this article with the other 14 travelers on my tour and look forward to future updates from Geisler and associates.
    Share:
    • Digg
    • del.icio.us
    • StumbleUpon
    • Facebook
    • Google Bookmarks
    • LinkedIn
    • Twitter
    • Print
    • email

    Related Topics

    , , , , ,