| correlation of single pixels [message #81971] |
Mon, 05 November 2012 01:51  |
haikoley
Messages: 5
Registered: November 2012
|
Junior Member |
|
|
Hi there,
I have two images showing the same scenery from different angles. What I want to find out is the pixel-offset of every(!) single pixel using correlation, but correlate and c_correlate only correlates two arrays and not a single value and an array. Is there something I can do?
Regards,
max
|
|
|
|
| Re: correlation of single pixels [message #82029 is a reply to message #81971] |
Mon, 12 November 2012 10:40  |
Klemen
Messages: 80
Registered: July 2009
|
Member |
|
|
Hi Max,
no x and y are in this case just the regression variables. See:
https://groups.google.com/forum/?fromgroups=#!topic/comp.lan g.idl-pvwave/F2E8cLePGsQ
Computing correlation using IDL function CORRELATE is ok, if you have to do it once. But here you have to do it within each area of interest a couple of times (considering set of possible offsets in x and y direction) so you have to use 4 FOR loops which will run slow!
Thus you should ask yourself first, how large is the data set you want to process. If you are sure that your offset are not larger than 2 pixels, then you might go for the slow version. That would be something like:
results_corr = make_array(...
results_offx = make_array(...
results_offy = make_array(...
FOR xall=xstart,xend do begin
FOR yall=ystart,yend do begin
search_area = data[xall-search_size:yall+search_size]
template = data[xall-template_size:yall+template_size]
FOR xoff=-search_size:search_size do begin
FOR yoff=-search_size:search_size do begin
data = (shift(search_area, xoff, yoff))[right indeice]
corr = correlate(data[*], template[*])
if corr gt results_corr[xall, yall] then begin
results_corr[xall, yall] = corr
results_offx[xall, yall] = xoff
results_offy[xall, yall] = yoff
endif
endfor
endfor
endfor
endfor
|
|
|
|
| Re: correlation of single pixels [message #82036 is a reply to message #81971] |
Mon, 12 November 2012 02:24 |
haikoley
Messages: 5
Registered: November 2012
|
Junior Member |
|
|
Hi Klemen,
Well, I'm quite new to IDL and that code still looks very complex to me (in fact I didn't expect it to be so complicated^^). I guess in your code x and y are the pixel offsets, is that correct?
Max
|
|
|
|
| Re: correlation of single pixels [message #84712 is a reply to message #82036] |
Mon, 12 November 2012 10:28 |
Klemen
Messages: 80
Registered: July 2009
|
Member |
|
|
Hi Max,
no x and y are in this case just the regression variables. See:
https://groups.google.com/forum/?fromgroups=#!topic/comp.lan g.idl-pvwave/F2E8cLePGsQ
Computing correlation using IDL function CORRELATE is ok, if you have to do it once. But here you have to do it within each area of interest a couple of times (considering set of possible offsets in x and y direction) so you have to use 4 FOR loops which will run slow!
Thus you should ask yourself first, how large is the data set you want to process. If you are sure that your offset are not larger than 2 pixels, then you might go for the slow version. That would be something like:
results = make_array(...
FOR xall=xstart,xend do begin
FOR yall=ystart,yend do begin
search_area = data[xall-search_size:yall+search_size]
template = data[xall-template_size:yall+template_size]
FOR xsearch=0,2*search_size+1 do begin
FOR ysearch=0,2*search_size+1 do begin
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
