Registration of TEM images

From MMVLWiki
(Difference between revisions)
Jump to: navigation, search
(Copied Manuel Boissenin's PMWiki page about TEM image registration)
 
m
 
(4 intermediate revisions by one user not shown)
Line 1: Line 1:
 
[[Image:ResImage15 16.png|thumb|300px|right|2D alignment of two TEM images]]
 
[[Image:ResImage15 16.png|thumb|300px|right|2D alignment of two TEM images]]
The Mimas registration algorithm has been adapted to register 2 TEM images of what I presume to be an indenter.
 
  
'''To sum up''':
+
When doing cross-correlation one typically uses a window function to avoid boundary effects. Instead of using a window function, one can use edge detection followed by cross-correlation for 2D alignment of images. The cross-correlation was performed using Fast Fourier Transform.
  
Preliminary results were not good. The inverse of the FFT is giving a maximum that doesn't correspond to the translation (at least the one we would naturally infer). Then I checked the high values and see that amongst them is the expected translation. My first idea was to threshold these high values and validate the hypothesis using another method (for instance a correlation measure or a difference one). But a muse wispered to my ear (nowadays we would probably talk about creative thinking) hey dude why don't you try to filter the image first as it seems that the lack of texture is a problem for this FFT method. So I tried and the result were improved. Still however the correspondence wasn't working time to time. So I did another attempt combining the original image with its edges and it seems it is working just fine.
+
=See Also=
 +
* [[Image:Mimasanim.gif|40px]] [[Mimas]]
  
The registration uses a windows that cut the original image to avoid taking the scale that disturb the registration method as there is not translation of it.
+
[[Category:Nanorobotics]]
 
+
[[User:Engmb|Engmb]]
+

Latest revision as of 12:53, 13 March 2009

2D alignment of two TEM images

When doing cross-correlation one typically uses a window function to avoid boundary effects. Instead of using a window function, one can use edge detection followed by cross-correlation for 2D alignment of images. The cross-correlation was performed using Fast Fourier Transform.

[edit] See Also

Personal tools
Namespaces
Variants
Actions
Navigation
Toolbox