active contours slows down half-way in a run

34 views
Skip to first unread message

Jingjing

unread,
May 17, 2016, 8:47:17 AM5/17/16
to Icy imaging
Dear Alexandre and Stephane,

I encountered a small problem when running active contours - it's very fast in the beginning when the cell contours are evolving, but once the contours reach the cell boundaries and appear "fixed", it suddenly slows down and takes ages to finish (actually I didn't wait for it to finish because it takes too long, and pressed the stop button prematurely instead to get the results...) I attached a screenshot when it halts. 

Is it normal or is there anything I can do to speed it up?

Many thanks,

Jingjing
active contours slows down in the middle of a run.png

Alexandre Dufour

unread,
May 17, 2016, 8:56:27 AM5/17/16
to icy-so...@googlegroups.com
Dear Jingjing,

Two ways to accelerate the process:
— Increase the “Evolution time step” parameter (beware: causes instabilities if too high)
— Increase the “Convergence criterion” parameter (to increase the detection rate of non-motile contours)
— Last resort (not recommended): reduce the maximum number of iterations.


Let me know if this helps
Alexandre


--
You received this message because you are subscribed to the Google Groups "Icy imaging" group.
To unsubscribe from this group and stop receiving emails from it, send an email to icy-software...@googlegroups.com.
To post to this group, send email to icy-so...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/icy-software/97fbe8ca-b06d-4aed-828d-a498b543a38e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<active contours slows down in the middle of a run.png>

Jingjing

unread,
May 17, 2016, 10:23:09 AM5/17/16
to Icy imaging
Dear Alexandre,

Thanks! Just tried all three - increasing the evolution time step gave larger objects, increasing the convergence criterion didn't show significant difference (I didn't bring it up to a crazy level though:)). and the least favourite "reducing iteration" worked! 

So I'll use the last resort while keeping the max iteration number as high as possible. 

Best,

Jingjing

ArpnSharma

unread,
Nov 2, 2016, 3:09:27 PM11/2/16
to Icy imaging
Hello,
Sorry if I should be posting somewhere else, but I found this post to be most relevant to my question. 
Background:
I want to track cell (2d+t). The cells are not evenly stained and when thresholding I get multiple "particles/blobs" for each cell. 
I have found active contours does the best job in identifying individual cells. It is "slow" when going through an image sequence of >150. I can try the suggestions you gave the OP concerning the speed. 
Question:
Is there a way to automate cell detection by active contours and then track the cells? Ideally, I would like to automate everything so that I can let it in my computer while I can be doing experiments. 

Thanks
Arpn 

Alexandre Dufour

unread,
Nov 14, 2016, 4:30:06 AM11/14/16
to icy-so...@googlegroups.com
Reply all
Reply to author
Forward
0 new messages