Re: Contribution: circles and hough transform
Hi,
FYI, I plan to work on the detection of ellipses quite soon. Stefan indicated to me a reference which seems fairly simple to implement.
Cheers, François.
Hi François,
I am trying out the ellipse hough transform as described in the following link: http://scikitimage.org/docs/dev/auto_examples/plot_circular_elliptical_houg...
The ellipse detection in the example takes about 15 seconds to give a result.
1. Is it possible to get the time taken for ellipse detection to under a second? 2. Are there any suggestions for improving the execution time?
Thanks, Austin
On Thursday, 7 March 2013 02:27:44 UTC+5:30, François wrote:
Hi,
FYI, I plan to work on the detection of ellipses quite soon. Stefan indicated to me a reference which seems fairly simple to implement.
Cheers, François.
 Is it possible to get the time taken for ellipse detection to under
a second? 2. Are there any suggestions for improving the execution time?
It's coded in cython, so it's pretty optimized and Johannes helped me on that. I think it's pretty well optimized.
You are trying to find a point is a space with 5 dimensions, so it takes times. The time directly depends on the number of (let's say black) points in your binary image. If you have a lot of points, you may try to reduce the quantity. I never tried so far, but it probably works if you take randomly a fraction of all your points. The peak in the 5D space will be smaller and perhaps wider, but depending on the fraction, you may find a compromise.
Best,
Thanks for the quick reply. I will give a shot at random fraction of all points but I have a feeling that it's not going to get it under one second. It makes sense why it's taking so much time.
Regards, Austin
On Wed, Jun 18, 2014 at 2:58 AM, François Boulogne fboulogne@sciunto.org wrote:
 Is it possible to get the time taken for ellipse detection to under
a second? 2. Are there any suggestions for improving the execution time?
It's coded in cython, so it's pretty optimized and Johannes helped me on that. I think it's pretty well optimized.
You are trying to find a point is a space with 5 dimensions, so it takes times. The time directly depends on the number of (let's say black) points in your binary image. If you have a lot of points, you may try to reduce the quantity. I never tried so far, but it probably works if you take randomly a fraction of all your points. The peak in the 5D space will be smaller and perhaps wider, but depending on the fraction, you may find a compromise.
Best,
 François Boulogne. http://www.sciunto.org GPG fingerprint: 25F6 C971 4875 A6C1 EDD1 75C8 1AA7 216E 32D5 F22F
 You received this message because you are subscribed to a topic in the Google Groups "scikitimage" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/scikitimage/yh_bI0hNIL8/unsubscribe. To unsubscribe from this group and all its topics, send an email to scikitimage+unsubscribe@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
participants (3)

Austin Chungath

François

François Boulogne