technochocolate

warning against subpixel rendering

Recently I was pointed to Thomas Maier’s site and stumbled across this article on using ‘subpixel hinting’ in Photoshop. This technique gives sharper image quality for almost all standard LCD screens in their standard orientation. So what’s the problem? The iPhone and the iPad. These new classes of device are made to view things in whichever orientation seems appropriate to the user. And with these platforms gaining steam all the time, statically subpixel-rendering assets becomes counter-productive.

The sharpness comes because the images take into consideration the minute differences in position of the red, green, and blue subpixels that make up each pixel. When we start physically altering the orientation of the screen (and thereby rotating the content that is displayed) our work to increase sharpness is not only lost, but in fact sharpness is actually worsened. In the case where the screen is rotated right or left, the subpixel rendering isn’t even working on the same axis as the subpixels anymore, and the even worse case, when it is upside down, the hinting works directly against the physical order of the subpixels.

On top of this issue, screen technology is still rapidly changing. Google’s Nexus One, for instance, has a screen with one green subpixel and one of either red or blue subpixels alternating in each adjacent pixel. Using subpixel rendering to generate assets bakes in assumptions about hardware that are no longer as valid as they once were. Point being, we should rely on hardware advancement to give us better image clarity and fidelity. One day the hardware might demand higher resolution images of us and offset the data for the subpixels dynamically, but we should avoid creating assets only fit for a single display system.

Update: This article was previously using the term ‘subpixel hinting’. It has come to my attention that what I am describing isn’t a correct usage of that term. I was simply responding using the language from Thomas’ article as I understood it. Font hinting is a term for aligning a font with the raster grid of a monitor so that it takes advantage of solid pixel edges where possible. What I’m cautioning against here is generating any art assets with a baked in presumption about subpixel arrangement to prevent improper display on screens of varying orientation. It appears that subpixel rendering is a more accurate term, so I’ve substituted it for subpixel hinting throughout the post. (Thanks to Typographica for the heads up.)