🚀 go-pugleaf

	Help,

	I'm working on a program to process images which have had
	bar code labels attached.  I need to find the bar code data
	in the midst of the scanned in images.  Once I find the piece
	of the image with the Bar Code I have no problem in orienting
	it and decoding it.  My problem is in finding the barcode.

	This appears to be a segmentation/feature extraction/roi
	problem of sorts.  I'm looking for input on an image feature
	I might use to search for a barcode and on a filter/processing
	for searching for that feature.  These images are bitonal (~200 dpi).

	Some things which come to mind are using a neighborhood thresholding.
	For instance taking a 5x5 region and mapping the percentage of pixels
	which are black ( x pixels / 25 pixels ) to a 256 level of gray image
	and doing blobs.  I've been getting enormous processing times in doing
	this.  A 7x7 filter calculating thresholds for every pixel in a 8.5x11
	(1700x2200 pixel) image took nearly five minutes.  And this was on
	a DECstation 5000/200.  I was hoping for < 20 seconds (maybe < 10).

	I've read some articles on texel (texture elements) and have thought
	about them or perhaps an FFT (because of the regular nature of bars).

	Does anyone have any ideas on some sort of processing which might
	help here?

	thanks,

	Mark Klamerus
	imaging systems consultant