function lct, first, second, x, y, gauss_sigma, Range = Range ;+ ; Name: LCT ; Purpose: ; Calculate local displacements between and cross-correlations ; of two successive images. ; Calling sequence: ; ; Result = lct (First_image, Second_image, Xpositions, Ypositions, $ ; Gauss_sigma, Range = Range) ; ; Inputs: ; First_image, Second_image: two images to be compared ; Xpositions, Ypositions: 1-d arrays of x- and y- pixel positions ; where displacements are to be determined. ; Gauss_sigma: Gaussian width of the apodizing window function ; ; Keyword input: ; ; Range: the maximum range of x- and y- displacements ; For the best results, displacement in each direction should ; be smaller than Range. ; (Range+1)*(Range+1) area is searched. ; (default value = 1 pixels) ; Output: ; Result : 3 by Npoint array ; Result(*, 0) x-displacements in pixels ; Result(*, 1) y-displacements in pixels ; Result(*, 2) cross-correlations (ranging from -1 to 1) ; These are set to zero at the pixels where ; determined displacements are bigger than Range. ; ;- on_error, 2 window_size = 2*fix(gauss_sigma*3.)+1 window = shift(exp(-0.5*(dist(window_size)/gauss_sigma)^2 ), window_size/2, window_size/2) window2 = sqrt(window) window = window/total(window) window2 = window2/total(window2) if n_elements(Range) eq 0 then Range = 1 hw_box = Range nb = 2*hw_box+1 delx = (indgen(nb)-hw_box)#replicate(1, nb) dely = replicate(1, nb)#(indgen(nb)-hw_box) delx0 = [ [-1, 0, 1], [-1, 0,1], [-1,0,1]] dely0 = [ [-1, -1, -1], [0, 0, 0], [1, 1, 1]] correlation = fltarr(2*hw_box+1, 2*hw_box+1) ; Preparation of Arrays required for Surface Fitting h = dblarr(9, 6, /nozero) h(*,0) = 1. h(*,1) = reform(delx0*2.) h(*,2) = reform(dely0*2.) h(*,3) = reform(delx0^2*4.) h(*,4) = reform(dely0^2*4.) h(*,5) = reform(delx0*dely0*4.) ht = transpose(h) hh = invert(ht#h) ; npoint = n_elements(x) d = fltarr(3, npoint) s=size(first) n = s(1) m = s(2) for k=0, npoint-1 do begin if k mod 500 eq 0 and k ne 0 then print,npoint-1,k xc = x(k) yc = y(k) x1 = xc - window_size/2 x2 = xc + window_size/2 y1 = yc - window_size/2 y2 = yc + window_size/2 inside = (x1-hw_box) ge 0 and (x2+hw_box) lt n and (y1-hw_box) ge 0 and (y2+hw_box) lt m if inside then begin for i=0, nb-1 do for j=0, nb-1 do begin dx = delx(i,j) dy = dely(i,j) first_sub = first(x1-dx:x2-dx,y1-dy:y2-dy ) second_sub = second(x1+dx:x2+dx, y1+dy:y2+dy) first_av = total(first_sub*window2) second_av = total(second_sub*window2) first_std = sqrt((total(first_sub^2*window) - first_av^2)>0.) second_std = sqrt((total(second_sub^2*window) - second_av^2)>0.) if (first_std*second_std) ne 0 then begin correlation(i,j) = (total(first_sub*second_sub*window) - first_av*second_av)/(first_std*second_std) endif else begin ;print, 'wrong1',first_std ;print, 'wrong2',second_std endelse endfor if nb gt 3 then begin ss= where(correlation eq max(correlation)) ic = ss(0) mod nb jc = ss(0) / nb endif else begin ic = 1 jc = 1 endelse if (ic-1) ge 0 and (ic+1) lt nb and $ (jc-1) ge 0 and (jc+1) lt nb then begin a = hh#(ht#reform(correlation(ic-1:ic+1, jc-1:jc+1),9)) peak = float(invert([[2*a(3), a(5)], [a(5), 2*a(4)]]) # [-a(1), -a(2)]) d(0, k) = peak(0)+delx(ic, jc)*2 d(1, k) = peak(1)+dely(ic, jc)*2 d(2, k) = float( a(0)+a(1)*peak(0)+ a(2)*peak(1) + a(3)*peak(0)^2+a(4)*peak(1)^2 + a(5)*peak(0)*peak(1)) if (abs(d(0,k)) gt range) or (abs(d(1,k)) gt range) then d(2,k)=0. endif endif endfor return, d end