PetaPixel

Super-Resolution From a Single Image

Super-Resolution From a Single Image” is an interesting research page by computer scientists over at the Weizmann Institute of Science in Israel. It details the group’s efforts to create sharp enlargements of small photographs, and offers comparisons between their algorithm and other popular ones being used and researched (e.g. nearest neighbor, bi-cubic). The large image of the baby seen above was created from the tiny image on the left. See if you can create something more useable using Photoshop.

Super-Resolution From a Single Image (via MetaFilter)


 
 
  • http://www.facebook.com/people/Jens-Hamada/100003300254445 Jens Hamada

    oh and they should try to sharpen the bicubic results.
    their examples look as if they have not sharpened them at all.
    as jeff schewe teaches you.. you have to sharpen when enlarging in photoshop using bicubic!!!

    try a simple highpass sharpening on their bicubic examples.
    i think you can come pretty close.

    ps: the comenting function acts weird.

  • http://www.facebook.com/profile.php?id=193303596 Zachary Larsen

    The only one that really showed a real improvement over methods that are already used was the image of the eyechart. In the example, most of the letters on the bottom row were actually legible using their algorithm. Very simple, high contrast examples were decent. Anything with fine detail just looked smudgy and oversharpened.

  • wickerprints

    After looking at the examples, I felt that the method of Kim, et al. was actually superior to that proposed by the authors. The baby picture shown above has too strong of a ringing artifact at high-contrast boundaries, which Kim’s method avoids. However, because of the limited number of comparisons of their method against others, it’s hard to say how the authors’ method stacks up against the others for a wider array of image sources.

    I think the best photographic application for such super-resolution algorithms would be minor image rescaling for the purpose of making prints–something at scales not exceeding around 200%. In other words, it would be useful for extracting better print output from an already good source. It would be an improvement over using bicubic interpolation, but only if one could avoid the creation of conspicuous artifacts. I don’t think it would be suitable for image enlargement for online display.

  • Nancy

    True. I love ononesoftware’s Perfect Resizing/