STANDARD HOUGH TRANSFORM TO DETECT ELLIPSE AND CIRCLES

i'm doing some final project in my campus, my topic is to analyze the difference between ellipse and circle detection using STANDARD HOUGH TRANSFORM now my problem is, i want to reconstruct the ellipse or the circle i've detect with hough transform, but i dont know how..

can anybody help me with this: .reconstruct the ellipse or the circle .is feature extraction needed for this kind of project? .how can i detect ellipse or circle with any noise above the object, for example a pen i put above the circle object..

here's the sample of a circle i want to detect : http://i1306.photobucket.com/albums/s578/andripranatakusuma/image12.jpg

here's the sample of a ellipse i want to detect : http://i1306.photobucket.com/albums/s578/andripranatakusuma/image15.jpg

i appreciate any kind of help, thank you :)

[Walter Roberson]

What error messages did you encounter when you implemented your code?

http://www.mathworks.com/matlabcentral/answers/6200-tutorial-how-to-ask-a-question-on-answers-and-get-a-fast-answer

[Andri Pranata]

with my codes, for now i can see the difference between ellipse and circle by measuring the standard deviation of the HOUGH map, though it was my own codes, but if i want to detect more than 1 object, my codes will not effective.. any suggestion?

here's my codes, i insert this code, when i've got the Standard Hough coordinater (rho theta) after detecting ellipse or circle.

==========================

capt = getframe(gca);
citra_capture = capt.cdata;
figure; imshow(citra_capture);

citra_capture = rgb2gray(citra_capture);
figure; imshow(citra_capture);

level = graythresh(citra_capture);
citra_capture_bw = im2bw(citra_capture,level);
figure; imshow(citra_capture_bw);

[jml_bar jml_kol] = size(citra_capture_bw);

%lebar = [];
for i = 1:jml_kol
    atas = find(citra_capture_bw(:,i) > 0, 1, 'first');
    bawah = find(citra_capture_bw(:,i) > 0, 1, 'last');
    if (~isempty(atas) && ~isempty(bawah))
        lebar(i) = bawah - atas;
    end
end
nonzero=find(lebar);
lebarnonzero=lebar(nonzero);
figure;plot(lebarnonzero);
lebar_rata = mean(lebarnonzero)
lebar_std = std(lebarnonzero)

Answer 1

[Harshit]

A simple solution is for say a circle of radius r. Pick up 10000 (any large number ) values between -r and r and feed them in variable x. Get y using circle equation. plot(x,y) and take a reflection along x axis.

[Andri Pranata]

when i can applied it? is it after the hough transform? before? or thats not using the hough transform for detection? any way thanx man :)