Humanist Discussion Group, Vol. 14, No. 803.
Centre for Computing in the Humanities, King's College London
<http://www.princeton.edu/~mccarty/humanist/>
<http://www.kcl.ac.uk/humanities/cch/humanist/>
Date: Wed, 18 Apr 2001 10:17:50 +0100
From: "John Unsworth" <jmu2m@virginia.edu>
Subject: RE: 14.0801 visual pattern recognition?
Willard,
A couple of pointers in response to your question about software's
capability in visual pattern recognition:
Blobworld:
http://elib.cs.berkeley.edu/photos/blobworld/
["Blobworld is a system for content-based image retrieval. By automatically
segmenting each image into regions which roughly correspond to objects or
parts of objects, we allow users to query for photographs based on the
objects they contain."]
Photobook:
http://www-white.media.mit.edu/vismod/demos/photobook/
["Photobook is a tool for performing queries on image databases based on
image content. It works by comparing features associated with images, not
the images themselves. These features are in turn the parameter values of
particular models fitted to each image. These models are commonly color,
texture, and shape, though Photobook will work with features from any model.
Features are compared using one out of a library of matching algorithms that
Photobook provides. In version 5,
these include euclidean, mahalanobis, divergence, vector space angle,
histogram, Fourier peak, and wavelet tree distances, as well as any linear
combination of these. Version 6 allows user-defined matching algorithms via
dynamic code loading."]
McGann, "Imagining What You Don't Know" (esp. section IV):
http://www.iath.virginia.edu/~jjm2f/chum.html
["There are critical opportunities to be exploited in the random use of
these kinds of deformation. For instance: take a given set of paintings and
run them through a series of deformations--let us say, a set of Rossetti's
paintings, or a set of works chosen from a known corpus of Pre-Raphaelite,
Aesthetic, and Symbolist art. Every time an altered image is generated that
is judged interesting or revealing in some way, save the settings by which
the image was produced. Then use those settings randomly on all the pictures
in the chosen set and compare the results."]
The bottom line is that software still isn't especially good at shape or
pattern recognition, but as the first two examples show, it's sometimes
better than you would expect, and as the last example suggests, even
software that can't *recognize* visual patterns might be able to *reveal*
them, if applied intelligently and systematically by a human being.
John Unsworth
This archive was generated by hypermail 2b30 : Wed Apr 18 2001 - 05:32:51 EDT