How to ellipse fitting the object

how to ellipse fitting the object?

http://snag.gy/oSz2X.jpg http://snag.gy/UBlgG.jpg http://snag.gy/e8mJG.jpg http://snag.gy/sI69L.jpg http://snag.gy/Rz8dB.jpg

[Image Analyst]

Background: Leon has already seen Matt J's code for fitting an ellipse: http://www.mathworks.com/matlabcentral/answers/53525#answer_65140

[Matt J]

This newer version is better:

https://www.mathworks.com/matlabcentral/fileexchange/87584-object-oriented-tools-for-fitting-conics-and-quadrics

Answer 1

[Image Analyst]

regionprops() can do that. It give the major axis, minor axis, and orientation for each blob in the image.

You may also want to see this paper: http://www.ecse.rpi.edu/homepages/qji/Papers/ellipse_det_icpr02.pdf

[L]

Hello,

This is an interesting topic. I have found few comments from you related to this topic. I have few questions since you are an expert (your comments are always very valuable).

1. regionprops() measures the parameters of one ellipse, or the parameters of all ellipses in the image?
2. What should be done if there are multiple ellipses with various dimensions? Can regionprops() handle that option?
3. Also, is it possible to modify the imfindcircles() to detect ellipses? Specially multiple ellipses, maybe attacked with noise too? This is a newer function in Matlab, it is pretty clear for circle detecion, but for the ellipses?

[Image Analyst]

1. regionprops() measures only the angle, centroid, and major and minor axis lengths. It does not give a full (x,y) equation.
2. regionprops() can handle multiple blobs in an image. You can return the results in a structure array (the default) or a table (if you specify the 'table' input option).
3. I don't know if it's possible to modify imfindcircles() to find ellipses. You can try to program up the attached paper if you want to find ellipses.

Also see this answer:

https://www.mathworks.com/matlabcentral/answers/53525-fitting-an-ellipse-to-my-binary-iris-image#answer_65140

and my attached demo.

[L]

Thank you.

Also, is it possible to use the imfindcircles() as an texture detector? Something like a Gabor filter, to highlight the curve-shaped textures?

I saw the paper that you have attached. But, what happens if the image is noisy? The method in the paper could be ssensitive in noise.

This is the reason I would try to detect the texutres first. Later, maybe the detection method would be combined with texture analysis.