reduce time taken for ellipse detection

[Austin Chungath]

Hi,

I am trying out the ellipse hough transform as described in the following link:

http://scikit-image.org/docs/dev/auto_examples/plot_circular_elliptical_hough_transform.html#example-plot-circular-elliptical-hough-transform-py

The ellipse detection in the example above with the coffee mug takes about 15 seconds to give a result.

Unfortunately, I need to be able to at least do this in under a second since I am working on a video stream.

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?

3. Are there any other library/package that can do ellipse detection better?

4. When I use my own image instead of the coffee mug - I get an error :Buffer and memoryview are not contiguous in the same dimension

I use the following code for loading the image the rest of the code is the same from the above example.

image_rgb = io.imread('ellipse.png') #instead of  image_rgb = data.coffee()[0:220, 160:420]

Obviously I must be doing something wrong.

 

I had mostly been working with OpenCV so I am a bit new to skimage's way of doing things. Please feel free to assume that I am a noob and any help is much appreciated.

Thanks,

Austin

[Dan Farmer]

I personally don't think I'd use scikit image for video work. Also the hough transform is going to be slow for video work. I'd suggest sticking with opencv and using MSER

See this example https://github.com/Itseez/opencv/blob/master/samples/c/mser_sample.cpp

--
You received this message because you are subscribed to the Google Groups "scikit-image" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scikit-image...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Stéfan van der Walt]

Hi Austin

On Tue, Jun 17, 2014 at 8:47 PM, Austin Chungath <aust...@gmail.com> wrote:
> I am trying out the ellipse hough transform as described in the following
> link:
> http://scikit-image.org/docs/dev/auto_examples/plot_circular_elliptical_hough_transform.html#example-plot-circular-elliptical-hough-transform-py
>
> The ellipse detection in the example above with the coffee mug takes about
> 15 seconds to give a result.
> Unfortunately, I need to be able to at least do this in under a second since
> I am working on a video stream.

This algorithm does a type of brute-force search, so I don't think it
will ever be fast enough. A probabilistic approach (as in the
probabilistic Hough line transform we implemented) would be better.
You can also scan the mailing list for a discussion we had a while ago
on fast detection of circles, which may be extended to ellipses.

Dan's suggestion of looking at OpenCV is also a good one; I am not
familiar with all the algorithms they provide, but it will undoubtedly
be fast.

Regards
Stéfan