Results 1 to 2 of 2

Thread: Scaling algorithm bug, found in most scaling algorithms

  1. #1

    Scaling algorithm bug, found in most scaling algorithms

    Interesting article :

    Check the main article, with a detailed explanation, by following the link from the slashdot story .

    To sum it up, it's not a very significant error visually, only under very specific circumstances (or it would be noticed sooner :P) but it's visible even in common images. If I understood the simplified version correctly, scaling down doesn't take gamma into account. Averaging a black and a white pixel results in a 50% grey pixel which visually has not 50% the luminance of white. For a 2.2 gamma, the correct average of black and white would be something like 73% grey.

    Very interesting, especially since, if it's indeed a bug, it's gone unnoticed by giants like photoshop for quite some time .

  2. #2
    Join Date
    Mar 2006
    Ambient Design
    Generally, even though it's not visually perfect, we have to keep the numerical purity of the pixel data, and post-apply gamma-correction. Otherwise repeated transformations are going to have other visual artifacts where the gamma appears to be artificially blown out.

    Also as developers we often have to make some compromises with perfect numerical accuracy for the sake of usability. For example while I could easily use 64-bit double-precision floating point maths for every channel operation, which would be visually flawless, it would make the application impossible to use and very slow.

    In the real world, application development is a juggling act of performance versus precision.
    AndyRage's mantra for graphics engine code:
    "Sure - how hard can it be?"

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts