Hi IDLers,
To transform vector info into raster images I am using the very usefull
IDLanROI object, with the ->ComputeMask() function.
Unfortunately, ComputeMask thinks "pixel", I think "grind point". Here
is a very simple program that illustrates what I mean. I programmed a
"MASK_RULE=3" if you want, that computes a mask following the rule
"center point of the pixel is included".
Here the code:
pro roimaskrules, npix
mask = BYTARR(npix, npix)
; Polygon vertices
x = [0.6,2.6,2.6,0.6,0.6]
y = [0.6,0.6,2.6,2.6,0.6]
roi = OBJ_NEW('IDLanROI', x, y)
t0 = SYSTIME(/SECONDS)
mask_0 = roi->ComputeMask(MASK_IN=mask*0B, MASK_RULE=0)
print, 'Time 0: ', SYSTIME(/SECONDS)-t0
t0 = SYSTIME(/SECONDS)
mask_1 = roi->ComputeMask(MASK_IN=mask*0B, MASK_RULE=1)
print, 'Time 1: ', SYSTIME(/SECONDS)-t0
t0 = SYSTIME(/SECONDS)
mask_2 = roi->ComputeMask(MASK_IN=mask*0B, MASK_RULE=2)
print, 'Time 2: ', SYSTIME(/SECONDS)-t0
; The "center point method"
t0 = SYSTIME(/SECONDS)
i = INDGEN(npix) # (LONARR(npix) + 1)
j = (LONARR(npix) + 1) # INDGEN(npix)
mask_3 = mask
cont = roi->ContainsPoints(i, j)
p_in = where(cont ge 1, cnt_in)
if cnt_in ne 0 then mask_3[p_in] = 255B
print, 'Time 3: ', SYSTIME(/SECONDS)-t0
if npix le 4 then begin ;print results to show what we want
print, 'mask_0: Boundary only.'
print, mask_0
print, 'mask_1: Interior only'
print, mask_1
print, 'mask_2: Boundary + Interior'
print, mask_2
print, 'mask_3: Contains centerpoint'
print, mask_3
endif
OBJ_DESTROY, roi
end
In this program, Mask_3 (and only mask_3) does what I want. For example
with a 4x4 array:
IDL> roimaskrules, 4
Time 0: 4.0054321e-05
Time 1: 3.0994415e-05
Time 2: 2.1934509e-05
Time 3: 6.3180923e-05
mask_0: Boundary only.
0 0 0 0
0 255 255 255
0 255 0 255
0 255 255 255
mask_1: Interior only
0 0 0 0
0 0 0 0
0 0 255 0
0 0 0 0
mask_2: Boundary + Interior
0 0 0 0
0 255 255 255
0 255 255 255
0 255 255 255
mask_3: Contains centerpoint
0 0 0 0
0 255 255 0
0 255 255 0
0 0 0 0
So, that works. But the problem is that if I want to do it for large
array this becomes VERY slow. Don't even try with large number of
polygons ;-):
IDL> roimaskrules, 4000
Time 0: 0.027777910
Time 1: 0.027390003
Time 2: 0.018553019
Time 3: 6.7377200
I come to the point: how can I make it faster? I know that testing each
point is the problem here, but I don't understand the real difference
between "my" MASK_RULE and IDL's MASK_RULE (why would IDL not implement it?)
Thanks for your help,
Fabz
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
