Quote:
Originally Posted by hawhill
*cough* errm, yes, would be a "nice to have", indeed. However - I think this would only make sense with a much more elaborated dithering algorithm. A simple ordered dither will make every soooo much more fuzzy... In fact, I was reading up on dithering algorithms because of the PDF reader app (which has priority, because it already has real practical use for me). It doesn't dither to grayscale, making pixel images in PDFs look a bit too "flat" (although my english vocabulary isn't small anymore, I'm lost in how to describe those big areas of a single grayscale tone with hard borders to the neighboring areas). So, stopping to brabble in my non-existent beard here - I would really love to have a fast error diffusion dither. I think it could work well for the VNC images, too (and of course video&games). The problem mainly is the "fast" part.
|
Yeah, dithering is not good for text, and VNC is MOSTLY used for text (the icons are generally redundant). But in the VNC demo video they showed a TV news program playing on the host PC desktop, mirrored to the kindle. In that special case, dithering would be nice. Now, if you want to get complicated, you could dither ONLY the parts of the image that are changing rapidly, and use grayscale for the parts that are static text.
I played with error diffusion a bit, but I personally liked an 8x8 ordered dither a LOT better on the eink displays. I memorized the Bayer ordered-dither matrix decades ago, and I can still type it out on demand from memory.
P.S. Those "big areas of a single grayscale tone with hard borders to the neighboring areas" are called "image segments", "blobs", "regions", "patches" (or other names). One way to find them is by using the "mexhat" function (similar to the one used in my algorithmic art scripts), but I think you would probably get a bigger kick from image segmentation using the "Hessian Blob Detector": http://en.wikipedia.org/wiki/Hessian_matrix.
