Close Contours of Image Segments

Hello

Working on a segmentation project a have fell upon a serious problem. Sometimes when my segmentation algorithm runs do not produces close contour boundaries, as it considers that the edge of the image is already a boundary but does not creates a boundary there. Is there a way to introduce this information to the segmented image ?

My problem is more complicated as I need to process one by one all the image segments so I only need the part of the boundary that creates a close contour with the segment each time.

Every boundary is exported using bwboundaries() and saved in a binary image (here occurs the problem)

Please if you have any idea how to overcome this problem let me know.

Thank you

Answer 1

[Image Analyst]

I've never heard of that, but you can try padarray() so that your object edges that are on the edges of the image are no longer on the edges of the image.

[Dimitris M]

Thank you for your answer but maybe I was not very clear, I need to keep the boundary.

For example look this image:

Image Boundaries

http://i.imgur.com/uB2jl.png

The image boundaries are not close contours so when I use bwboundaries() to get the segments.

Do you think that by introducing the perimeter in the image boundary would improve the segment extraction ??? I was thinking to make a dilation and then subtract the original image so I can have the borders.

Thank you in advance

[Image Analyst]

Let's back up. What do you really want to do? Can you post your grayscale image, and post your binary image? Then tell me what you want to find. Because I don't see any reason at all to find boundaries of things that are already 1 pixel wide lines.

[Dimitris M]

1)So this is my Image:

http://imm.io/qM7F

2) This is the segmentation result (on the RGB)

http://imm.io/qM89

3) This is the superimposed

http://imm.io/qM8v

As you can see the boundaries are not closed (creating polygons) as there are some gaps in the boundaries. I am know trying to export each boundary individually in an image using bwboundaries() and plot the locations on a black image. In this way I want to create polygons and calculate parameters for them.

The problem is that if I do not have closed polygons I can not calculate parameters for them. I think by superimposing the original image borders I can work this out as then I will have polygons in every case.

Am I clear or should I provide you more information ???

Thank you again for your time and patience !