X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: f996b,7f7d1aa9b39cc9f5
X-Google-Attributes: gidf996b,public
X-Google-ArrivalTime: 2001-04-27 01:38:22 PST
Path: 
 newsfeed.google.com!newsfeed.stanford.edu!news.tele.dk!193.174.75.178!news-fra1.dfn.de!news-fra.pop.de!schlund.de!rz.uni-karlsruhe.de!news.rz.uni-karlsruhe.de!not-for-mail
From: Markus Gebhard <ukgh@rz.uni-karlsruhe.de>
Newsgroups: alt.ascii-art
Subject: Re: image 2 ascii logo
Date: Fri, 27 Apr 2001 10:38:28 +0200
Organization: University of Karlsruhe, Germany
Lines: 42
Message-ID: <3AE93004.EF52674F@rz.uni-karlsruhe.de>
References: <eli$0104251931@qz.little-neck.ny.us>
 <3AE7C44F.3F72B3D4@rz.uni-karlsruhe.de> <eli$0104261621@qz.little-neck.ny.us>
NNTP-Posting-Host: wn4-jarjar.wn4.uni-karlsruhe.de
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Trace: news.rz.uni-karlsruhe.de 988360701 4961 172.20.12.141
X-Complaints-To: usenet@rz.uni-karlsruhe.de
X-Mailer: Mozilla 4.77 [en] (Win98; U)
X-Accept-Language: en
Xref: newsfeed.google.com alt.ascii-art:5490

Eli the Bearded wrote:

> Will you be sharing your algorithms? I don't care much about the
> actual code, just the methods behind it.

I consider releasing the source code of the image2ascii part of Jave
later this year.
But the algorithms in there are quite easy, I have already discussed
them in another thread in this group:
"Re: Gwyneth Paltrow - Jave image->ascii conversion .. and discussion of
algorithms" 

I just take 4 (or 1 - depending on the algorithm one uses) pixels from
the origional image and try to find the best fit in a table. In this
table (almost) all ascii characters have asigned values for their grey
shade in each quarter.
So there is only one tricky part: Find a good table with those gray
values. I have written a program to create those tables and I was very
surprised that:
a) the results of the 2x2 conversion is that good
and
b) when you take 3x3 pixels it does not work any longer.

So my algorithms are based on gray shades which only seem to work for up
to a grid of 2x2, while other algorithms are based on a black/white
decission. E.g. you can take a 3x3 grid and asign black/white pixels for
each region to each ascii character:
+---+
| # |
|###|
|# #|
+---+

For future enhancements in Jave's image2ascii tool, I consider writing
additional algorithms by
a) trying grey shades in a 2x3 or 3x2 raster
b) using the black/white methode with some ideas I have

So now that I have put my cards on the table, what are your ideas and
experiences?

Markus