NAME:
	xyoff
 PURPOSE:
	Calculates the correlation function between two images using a box of
	size (bx,by) surrounding a point in the image. By default this point
	is the center of the image, (nx/2,ny/2). The routine calculates the
	offsets by finding the maximum of the correlation function.
 CATEGORY:
	Image processing
 CALLING SEQUENCE:
	s = xyoff(a,b,bx,by, hx,hy,fna,wdow=wdow,mask=mask,/sobel,ccf=ccf,/quiet)
 INPUTS:
	a	2D reference image.
	b	2D image that you want to track with respect to a
	bx	size of the box in the x-dimension that you wish
		to use to generate the cross correlation function
	by	size of the box in the y-dimension that you wish
		to use to generate the cross correlation function.
	hx	x position in the image that corresponds to the center
		of the box. (Optional)
	hy	y position in the image that corresponds to the center
		of the box. (Optional)
	fna	fourier transform of the 256x256  (Optional)
		subtended image from the center of a.
 OPTIONAL INPUTS:
	wdow=wdow	2-d apodization window.  (Optional; default=1)
	mask=mask	2-d mask to be applied in fourier space. (Optional;default=1)
	/sobel		If specified, then SOBEL keyword causes this routine
			to track on the gradient of the image instead of the
			image itself.  (It uses a Sobel edge enhancement operator)
	/quiet		if set, suppress printing of results on screen
 OUTPUTS:
	s =	A vector containing the x and y offsets that should be applied
		to an image to maximize the cross correlation.
	s[0] = The x offset.
	s[1] = The y offset.
	s[2] = max value of the correlation
	s[3] = cross correlation at 0 shift
	s[4] = maximum of the real part of absolute value of cross correlation
 INCLUDE:
	@compile_opt.pro	; On error, return to caller
 CALLS: ***
	InitVar, MEAN, max_pos
 PROCEDURE:
	The executing program is xyoff.pro.  It performs a 2D
	cross-correlation, yielding the lag of maximum correlation, the
	coefficient of maximum correlation and the coefficient at lag 0.  In
	practice, I usually used lag 0.  Within the limits of floating point
	computation, the cross correlation at 0 shift of an image and itself is
	1.0 and the shifts are 0.0 and 0.0 as would be expected.  xyoff.pro
	calls max_pos.pro to find the maximum position within an array...the
	cross-correlation function in this case.
 MODIFICATION HISTORY:
	H. Cohl,   7 Jun, 1991 --- Initial programming.
	H. Cohl,  13 Dec, 1991 --- Added SOBEL keyword.
	H. Cohl,  13 Feb, 1992 --- Made it so that you can enter in hx or hy.
	H. Cohl,  25 Mar, 1992 --- Made documentation more understandable.
	T. Henry   December, 1993 ---  some changes made to size