Offset b/w Bayesian and single-emitter-fitting results

Bayesian analysis of blinking and bleaching

Offset b/w Bayesian and single-emitter-fitting results

Postby hooying » Sun Apr 06, 2014 4:36 am

Hi,

I performed Bayesian analysis and single-emitter-fitting analysis using a single data file (128 by 128 px using submasks), which contains both bleach-and-blink frames and single-molecule frames. When I overlaid the two results, I found a slight offset between the two (blue: Bayesian; red: single-emitter fitting). I wonder what could have caused that. Thanks, -Ying
Attachments
Screen Shot 2014-04-05 at 9.29.59 PM.png
Screen Shot 2014-04-05 at 9.29.59 PM.png (110.99 KiB) Viewed 19136 times
hooying
 
Posts: 6
Joined: Tue May 14, 2013 10:03 pm

Re: Offset b/w Bayesian and single-emitter-fitting results

Postby edrosten » Thu Apr 10, 2014 11:31 am

hooying wrote:Hi,

I performed Bayesian analysis and single-emitter-fitting analysis using a single data file (128 by 128 px using submasks), which contains both bleach-and-blink frames and single-molecule frames. When I overlaid the two results, I found a slight offset between the two (blue: Bayesian; red: single-emitter fitting). I wonder what could have caused that. Thanks, -Ying


It's tough to say without more information. Do you have a scale bar for the image?

Probably the most common cause for offsets with this kind of data is different conventions with respect to coordinates: it's arbitrary whether the centre or corner of a pixel is at (0,0). This will lead to a diagonal offset exactly half a pixel. Does that explain the offset, or does that not fit?

-Ed
edrosten
 
Posts: 42
Joined: Wed Jul 18, 2012 2:54 pm

Re: Offset b/w Bayesian and single-emitter-fitting results

Postby hooying » Wed Apr 16, 2014 4:21 am

Thanks Ed. It appears adding half a pixel to x and y corrected for the offset. I am attaching a screenshot with the correction.

Ying
Attachments
Screen Shot 2014-04-15 at 9.20.10 PM.png
Screen Shot 2014-04-15 at 9.20.10 PM.png (72.3 KiB) Viewed 19114 times
hooying
 
Posts: 6
Joined: Tue May 14, 2013 10:03 pm

Re: Offset b/w Bayesian and single-emitter-fitting results

Postby hooying » Wed Apr 16, 2014 11:03 pm

Is it possible the offset is 1 instead of 0.5 pixel? I am attaching a picture with the Bayesian result offset by one pixel and the alignment appears to be better with the single-emitter fitting data. The x and y axis is labeled by the pixel number and the pixel size is 160 nm. Ying
Attachments
Screen Shot 2014-04-16 at 3.57.05 PM.png
Screen Shot 2014-04-16 at 3.57.05 PM.png (116.46 KiB) Viewed 19113 times
hooying
 
Posts: 6
Joined: Tue May 14, 2013 10:03 pm

Re: Offset b/w Bayesian and single-emitter-fitting results

Postby edrosten » Fri Apr 18, 2014 3:42 pm

hooying wrote:Is it possible the offset is 1 instead of 0.5 pixel?


Certainly: a 1 pixel offset would be caused by indexing differences. It's common to consider the first pixel in an image to be either (0,0) or (1,1). Using (0,0) is more common in code from C-derived languages (C, C++, Java) and (1,1) is more common in FORTRAN derived ones (Fortran, Matlab, Octave).

Much like with the 0.5 pixel offset, the choice is completely arbitrary.

-Ed
edrosten
 
Posts: 42
Joined: Wed Jul 18, 2012 2:54 pm

Re: Offset b/w Bayesian and single-emitter-fitting results

Postby FrankGarza » Wed Dec 31, 2014 7:22 am

It's hard to state with out more information. Have the level club to the photograph?
One of the most typical result in with regard to offsets using these kinds of facts differs events regarding coordinates: it really is human judgements if thez centre or even corner of any pixel is at (0, 0). It will produce the diagonal balanced out precisely 50 % the pixel. Will of which reveal the particular balanced out, or even may of which not fit?
FrankGarza
 
Posts: 1
Joined: Wed Dec 31, 2014 7:20 am


Return to 3B

Who is online

Users browsing this forum: No registered users and 1 guest

cron