Connect pixels to form the boundary of an object

Hi,

I want to connect disjoint pixels to form the boundary of the object. What are the necessary commands/procedures to do so?

It is quite urgent. Thanks so much if anyone can answer it.

Answer 1

[Image Analyst]

You can use activecontour(), demo attached.

Or you can use edge linking. http://www.peterkovesi.com/matlabfns/#edgelink

Or maybe convhull(), bwconvhull(), contour(), bwboundaries(), bwperim(), etc. depending on what your image looks like.

[TY]

Hi Image Analyst,

Actually I am doing a project which the instructor asks me to conduct template matching by using a template image that is different from the object in the input image( Assume only one object in the image, i.e. 1 object 1 background). For example, he suggests me to use a circle (template) to match with a triangle (target object) by some proposed algorithms.

I try to use phase correlation (canny edge) to generate a correlation performance score and save the score as matrix. If it exceeds a certain score, the pixel is selected and they will be connected to form the boundary of the object (Segment the contour of the object).

What commands/ procedures will be better to use in the above case?

[Image Analyst]

You might try normalized cross correlation, done by normxcorr2(). See attached demo.

[Kimo Kalip]

Two questions:

1. Is this a literal recognition, where given the photo at subplot (2,2,2) will only recognize in that particular orientation in this particular image, or could you rotate the onion, or have two such onions in the photo and you'd still get matches?
2. How exactly does normalizing the original image work for you in subplot (2,2,3)? Like I don't understand how it tells you where the onion is.