6
$\begingroup$

Subpixel rendering is used most commonly to anti alias fonts. It works by leveraging the physical layout of the color components of a display to give geometry details to an image that are smaller than a pixel.

For instance, this font has a height of 3 pixels but can easily be read: enter image description here

from: http://www.sitepoint.com/two-teeny-tiny-fonts/

This is a monochromatic case, and i could see this possibly working in a greyscale case, but it seems like it always has to be used in situations where you have high contrast.

Is it possible to use sub pixel rendering in a full color situation, such as those you'd find while doing a 3d render of a scene?

I think that it must not be, since it won't be high contrast enough, but does anyone have examples to the contrary?

$\endgroup$

1 Answer 1

5
$\begingroup$

Here's an example of a downsampling filter that takes pixel geometry into account: Increasing image resolution on portable displays by subpixel rendering.

Image (a) below is an image downsampled using pixel-based downsampling. Image (b) is downsampled using direct subpixel-based downsampling, which (as far as I can tell) effectively downsamples the R, G, and B planes of the image independently. Image (c) is downsampled using diagonal direct subpixel-based downsampling, which uses a diagonal pattern to improve apparent resolution in both horizontal and vertical directions.

Figure 11 from the paper

You can see that the last two images look sharper, but also have the characteristic color fringing.

Here's the downsampling pattern used for the last image:

DDSD downsampling pattern

$\endgroup$
4
  • $\begingroup$ By displaying the images magnified, you entirely negate the benefit of using subpixels, so the comparison images aren't representative of the actual results. You should try posting the original-size images if possible. $\endgroup$
    – yuriks
    Commented Sep 4, 2015 at 5:48
  • 1
    $\begingroup$ @yuriks There's another smaller resolution that I can find (journals.cambridge.org/fulltext_content/SIP/SIP1/…) but it's not immediately apparent to me that that version is not minified or magnified. I do think it's educational to see the magnified version, since it lets you see the fringing and doesn't actually get rid of the sharpness relative to the regular downsampling. $\endgroup$ Commented Sep 4, 2015 at 5:57
  • 1
    $\begingroup$ @yuriks, problem with these kinds of images are that not all monitors have same subpixel sequence. And on mobile devices the orientation needs to change when the user turns the device. $\endgroup$
    – joojaa
    Commented Sep 4, 2015 at 6:21
  • $\begingroup$ @JohnCalsbeek A subpixel-rendered image has its benefits only at a certain resolution. Upscaling it replicates (tiles) its local subpixel patterns, e.g. a diagonal line turns into a diagonal zig-zag. That may look sharp, but it's not the original image contents. $\endgroup$
    – root
    Commented Oct 25 at 14:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.