Image processing and alignment

  • Follow


Hey all,,

I have a question with image alignment. I have two sets of height data of a surface at different temperatures from an experiment. The data set has both translation and rotation. The data is saved as matrix format. Now i need to align those data sets.. What I have done is, I  converted the matrix to gray format using mat2gray and found the edges of feature on the surface using canny filter. I found the angle by selecting a set of points in edge plot (logical data set) and fitted a line to it and then found the angle for that line. I have removed the rotation from both the data set using imrotate. Now the angle is removed I need to correct for translation in x and y. 
   Since I am doing the experiement at different temperatures some noise factor add in to the data set. I know I can use cpselect for manually selecting the contol points and then find the translation. Wondering if there is any automatic way of doing it? 

Thanks
Deepak
0
Reply Deepak 7/13/2010 4:43:05 PM

Forgot to add in this... Normalized Cross-Correlation isn't working well because of noise factor... 
0
Reply Deepak 7/13/2010 4:53:04 PM


"Deepak Sankar" <deepaksankar@yahoo.com> wrote in message <i1i5lg$1a2$1@fred.mathworks.com>...
> Forgot to add in this... Normalized Cross-Correlation isn't working well because of noise factor... 

Do a search on here for "phase correlation."  It can handle both rotations and translations as well as scaling.  Also it's fairly resilient to noise and intensity differences.  
0
Reply Sean 7/13/2010 6:05:08 PM

"Deepak Sankar" <deepaksankar@yahoo.com> wrote in message <i1i5lg$1a2$1@fred.mathworks.com>...
> Forgot to add in this... Normalized Cross-Correlation isn't working well because of noise factor... 

Also, see this submission:
http://www.mathworks.com/matlabcentral/fileexchange/18401-efficient-subpixel-image-registration-by-cross-correlation
0
Reply Sean 7/13/2010 6:06:10 PM

3 Replies
348 Views

(page loaded in 0.022 seconds)

Similiar Articles:













7/29/2012 4:38:36 PM


Reply: