Getting the circle boundary

Question

I have a logical circular mask of 0 and 1 in my matrix which looks the following way.

What is the fastest way to get just the outside boundary in another matrix ?

Essentially I have to scan for the first 1 from left and first 1 from right in each row if there are duplicate 1's in the row ( there will be just one 1 in the topmost , bottommost points ).... Can someone help me in finding a fast way to do this ?

Answer 1

You can use regionprops for that, here are a few examples that identify circles:

• link1
• link2
• link3

or if you're sure there's only one circle and no noise, I assume you can just find the bottom/top/left/right edge and work from that:

m = loadcirclefunction();
pix_left  = find(any(m,1),1,'first');
pix_right = find(any(m,1),1,'last');
pix_top   = find(any(m,2),1,'first');
pix_bottom= find(any(m,2),1,'last');

Answer 2

The other answers are good for finding general circles in images, but since you know that you're looking for a circle in a binary mask, bwmorph is probably your best bet.

I=imread('0ateM.png');
BW=im2bw(I);
BW2=bwmorph(BW,'endpoints');

Edit: As I mentioned in the comments, in order to enlarge the circle so that you're getting the 0 pixels just outside the original circle mask set to 1 and everything else set to 0, you can invert the original mask and then use bwmorph:

WB=-(BW-1);
WB2=bwmorph(WB,'endpoints');

This has the unfortunate side-effect that the border of the image is changed to 1's. Of course you can easily change this. For an mxn image:

WB2(1,:)=0; WB2(:,1)=0; WB2(:,n)=0; WB2(m,:)=0;

An alternative approach is to directly use a filter on the original image:

f=[1 1 1; 1 -9 1; 1 1 1];
G=filter2(f,BW);
BW2=im2bw(G);

This will achieve the same results as WB2 above without the white border problem. The im2bw call is needed because after the filter the values are no longer just 0 or 1, they range somewhere between -8 and 8 and we want the negative values to be 0 and the positive values to be 1.