Difference between revisions of "TrackMate Algorithms"

m (Draft)
 
Line 1: Line 1:
THIS IS A DRAFT WHERE SNIPPETS ARE LEFT
+
This page documents the current components of [[TrackMate]]. TrackMate has a modular design, meaning that it is made of different modules that each have a specific role. Developers can build their own module and re-used the other ones and the GUI to achieve a quick development. The module types are (in the order you meet them when executing the plugin):
 +
 
 +
# '''Spot segmenters'''. Taking your image data, they detect spots in them.
 +
# '''Spot feature calculators'''. Each spot can receive a wide range of features, calculated from their location, radius and the image data. For instance: max intensity in spot, rough morphology, etc... They are then used to filter out spurious spots and retain only good ones for the subsequent tracking step.
 +
# '''Diplayers'''. Display the segmentation and tracking results overlaid on your image data.
 +
# '''Spot trackers'''. Take the filtered spots and link them together to build tracks.
 +
# '''Track feature calculators'''. Like for spot feature calculators, but operate on whole track. Can be used to report track mean velocity, displacement, etc... They are also used to filter spurious tracks.
 +
# '''Actions'''. Miscellaneous actions you can take on the complete result of the tracking process. It can be used to copy the track overlay to another image, launch a 3D viewer, export the results to a simple format, generate a track stack, etc...
 +
 
 +
We describe here the best we can the current modules that are shipped with TrackMate.
 +
 
 +
 
 +
== Spot segmenters ==
 +
 
 +
Behind this barbaric name stand the part responsible for spot detection in your image. The algorithm currently implemented are very generic and naive. They will most likely fail for complicated case such as touching objects, very weak SNR, etc... The 3 of them present are all based on Laplacian of Gaussian filtering, which we describe below.
 +
 
 +
 
 +
 
 +
 
  
  

Revision as of 10:41, 15 June 2012

This page documents the current components of TrackMate. TrackMate has a modular design, meaning that it is made of different modules that each have a specific role. Developers can build their own module and re-used the other ones and the GUI to achieve a quick development. The module types are (in the order you meet them when executing the plugin):

  1. Spot segmenters. Taking your image data, they detect spots in them.
  2. Spot feature calculators. Each spot can receive a wide range of features, calculated from their location, radius and the image data. For instance: max intensity in spot, rough morphology, etc... They are then used to filter out spurious spots and retain only good ones for the subsequent tracking step.
  3. Diplayers. Display the segmentation and tracking results overlaid on your image data.
  4. Spot trackers. Take the filtered spots and link them together to build tracks.
  5. Track feature calculators. Like for spot feature calculators, but operate on whole track. Can be used to report track mean velocity, displacement, etc... They are also used to filter spurious tracks.
  6. Actions. Miscellaneous actions you can take on the complete result of the tracking process. It can be used to copy the track overlay to another image, launch a 3D viewer, export the results to a simple format, generate a track stack, etc...

We describe here the best we can the current modules that are shipped with TrackMate.


Spot segmenters

Behind this barbaric name stand the part responsible for spot detection in your image. The algorithm currently implemented are very generic and naive. They will most likely fail for complicated case such as touching objects, very weak SNR, etc... The 3 of them present are all based on Laplacian of Gaussian filtering, which we describe below.




Calculating linking costs

  1. The distance between the two spots D is calculated
  2. If the spots are separated by more than the max distance, the link is forbidden. If not,
  3. For each feature in the map, a penalty p is calculated as

 p = 3 \times W \times \frac{ | f_1-f_2|}{f_1+f_2}

where W is the factor associated to the feature in the map. This expression is such that: there is no penalty if the 2 feature values f1 and f2 are the same; that, with a factor of 1, the penalty if 1 is one value is the double of the other; the penalty is 2 if one is 5 times the other one. All penalties are summed, to form P = (1 + ∑ p ) The cost is set to the square of the product: C = ( D × P )²