X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: fd588,b95c078a91be9537
X-Google-Attributes: gidfd588,public
X-Google-Thread: f996b,b95c078a91be9537
X-Google-Attributes: gidf996b,public
From: iltzu@sci.fi (Ilmari Karonen)
Subject: Re: GIF conversions (was: Re: History)
Date: 1998/08/30
Message-ID: <6sc7t0$4og$1@tron.sci.fi>#1/1
X-Deja-AN: 386206843
Distribution: world
References: <903316543.458387@mnementh.southern.co.nz>
 <35DAC105.AE270BBC@spamfree.land> <903556665.490124@mnementh.southern.co.nz>
 <35DB58B8.690CA54A@spamfree.land>
Followup-To: alt.ascii-art,alt.ascii-art.animation
Organization: (dis)Order of the Holy Spoon (or whatever)
Newsgroups: alt.ascii-art,alt.ascii-art.animation

Matthew Thomas (mpt26@spamfree.land) wrote:
: One day, I promise, when I have enough time to fiddle around with such
: things, I will write the ultimate GIF-/JPG-/PNG-to-ASCII converter
: (yeah, right!) (... no, really!).
: I plan the following features:
: * optional edge-detection -- drawing of outlines will take precedence
:   over shading the areas inside them
: * as much or as little weight as you like will be able to be given to
:   the shape of the characters, rather than their darkness or lightness
:   (ie, anti-aliasing at the expense of exact shade, or vice versa).

Hmm. I have several half-finished bitmap->ascii routines lying around.
On uses a statistical method to find the best matching character for
each 3x3 greyscale block, another does a very fast mapping from an
"interleaved" b/w bitmap. Both have significant problems left.

The advice I would give is - if at all possible - to let the bitmap
rectangles determining each character overlap. This is especially
useful with line art, as it can give much smoother curves and gets
rid of the "borderline" problem. (A horizontal line giving ______
looks a lot worse in most converters if shifted down one pixel to
become       .)
       """"""
The only problem is that this tends to make the number of pixels per
character even larger, making a simple lookup table approach 
impractical. I have an idea involving binary trees, and another
using boolean masks, but I haven't had the time to try either.

-- 
Ilmari Karonen (iltzu@sci.fi)
http://www.sci.fi/~iltzu/