[Devel] Re: Comments on Microsoft ClearType technology

[Berthold K.P. Horn]

This was a popular topic subject of heated debate a year ago.  
Along the lines of yet more stuff that MS stole from Mac,
who stole it from Xerox PARC, who stole it from academia etc.

The short summary may be:

(1) a color display can and has been used to  get apparent higher resolution
for B/W images by using the color channels in combination;

(2) Such ideas are very old, of course. More recently, such methods were used 
in the early days of the Mac on CRT type displays using two channels, but such
methods are in several ways different from what is used in ClearType;

(3) The application to LCD is different in that the three (or four) sub-pixels
are in predictable locations, unlike CRT where there is no fixed relationship 
between the R G and B dots and the pixels addressed by the video board;
The methods used in ClearType do not help on a traditional CRT display
(despite what was said by MS people in a news conference);

(4) Different LCD color displays have different arrangements of the sub-pixels,
and the algorithm needs to take this into account. Unfortunately the
standard interface for displays does not feed back useful information
to the video board on such matters.  

(5) If the screen does not match the assumptions of the algorithm you get 
*lower* resolution and color fringes.  Some people have trouble picking the
right "polarity" just by looking at the screen, which perhaps tells you 
something
about how subtle the effect is!

(6) There are claims that you need to do clever things to avoid colored
fringes, but simple Floyd Steinberg error propagation methods work fine.
The lower resolution of the human visual system for color compared to
brightness helps hide any such effects in any case, even when nothing
special is done to suppress color fringes.

(7) The only place were it has been noted of significance is in electronic
books, which are seriously resolution starved.  MS is apparently bringing
out some electronic book devices with LCD displays that exploit this.

Regards, Berthold.

At 01:18 AM 2/1/2001 +0100, Werner LEMBERG wrote:
>> Essentially, this (patented) technology uses subpixel sampling based
>> on knowledge of how LCD displays simulate color (they do it quite
>> differently from CRT displays) to improve the display of fonts and
>> line art.
>
>I haven't investigated the patent issue further, but subpixel sampling
>isn't new at all -- a similar technique has been used for old Apple II
>computers, as discussed in
>
>  http://grc.com/cleartype.htm
>
>This page also demonstrates very well how it works. 
>
>
>    Werner

--
Berthold K.P. Horn  mailto:address@hidden  http://www.ai.mit.edu/people/bkph 
(MT)