NAME
	feature	- does feature extraction for preprocessed signatures

SYNOPSIS
	feature if=infile [tf=target output file] of=outputfile [-d] [n=] [t=] [-h]

DESCRIPTION
	Extracts a set of features for signature verification.  There are Global features, which analyze the entire image, and stroke features.
	In this algorithm, a stroke is a curve that does not change upward/downward direction.  i.e. if you start at the bottom pixel of a stroke, you can trace a path of all the pixels of the stroke by only using pixels above or on the same level as the current pixel.  Features are extracted for each stroke, giving features local to that stroke.  If the signature is too large, a stroke limit is imposed, and any strokes after a certain limit are neglected.

CONSTRAINTS
	The signature is assumed to have been preprocessed.  The 2 things preprocessing must enforce are
	1.  The image is binary, with values 0 and 1.
	2.  The image has been thinned to 1 pixel width.

OPTIONS
	There are 2 main modes of operation.  If the input file if has multiple frames, features will be extracted for each frame.  A single frame image can also be used.
	tf   -   If a multiple frame image is input, the target values(real/fake)  are put in the target file.
        -d   -   Debug mode
        n    -   Gives the number of frames in the input file (found with vls).  Default is 1
        t    -   If n=1, then instead of using a tf file, the taget value can be given as t
        -h   -   Puts the vnet header into the features vector.  If other classification methods are being used, leave this option out

AUTHOR(S)
	A.Mehler and U.Moszkowicz
