Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!swrinde!ihnp4.ucsd.edu!munnari.oz.au!cs.mu.OZ.AU!munta.cs.mu.OZ.AU!caburt
From: caburt@munta.cs.mu.OZ.AU (Craig Alexander BURTON)
Subject: 14th century music restorn. want to try SOFM..ideas?
Message-ID: <9508516.15920@mulga.cs.mu.OZ.AU>
Summary: Attempting a restoration of 1400c music bitmaps with a SOFM.
Keywords: SOFM 14th century music neural net graphics
Sender: news@cs.mu.OZ.AU (CS-Usenet)
Organization: Computer Science, University of Melbourne, Australia
Date: Sun, 26 Mar 1995 06:43:01 GMT
Lines: 36


	Hi folks,

	I am attempting a restoration of some badly microfiched
	14th century music sheets.

	They are all the same size, all bitmaps and look like
	a classic OCR job.  I would, however, like to avoid
	a complete OCR approach, and want to use a SOFM with
	a special training image.

	I have done this before -- used a SOFM as a sort of
	static compressor -- and got great image enhancement
	with a training image of concentric graded circles.

	It looks like the symbol set for the music is very
	small -- it is much simpler than modern music notation
	and I will try to construct a training image of well-
	formed notes, staves etc.  I hope to overcome variations
	in position of these notes by supplying _all_ permutations
	of notes within 8x8 codewords (64 per note type).

	Some of the notes are very badly degraded, with only
	an outer rim of pixels remaining from a solid figure.
	I remember seeing some amazing reconstruction in Zurada
	with BAMs.  Is this possibly a better idea??

	Please mail me with any interest, as I would greatly
	appreciate any advice on this experiment.

	Craig Burton
-- 
--------------------------------------------------------------------
Mr. Craig Alexander BURTON        caburt@cs.mu.oz.au 
Dept. of Computer Science         +61 3 344 9150 or
Vision Group                      489 4386 or 479 2398 
