Image Processing Question - Correlating Watershed Segmented Image with Original Image

Hello,

I have created a segmented image of an original image (both 256x256 pixels) with a watershed algorithm. In the segmented image, the lines (or segments) correspond to "grain boundaries" in my original image, which are significant features. I need to determine the gray values in the original image at these grain boundaries.

What I have done thus far is create a text file for the original image and for the segmented image which are both 256x256 lists of gray values, 0-255. The text file for the segmented image only has 0s (black in the image) and 255s (white in the image). The pixels with 255s are the places where the grain boundaries are in the original image. Therefore I need a code that returns the gray values in the original image that correspond to the 255 values in the segmented image. Do you have an idea of how to accomplish this?

Or, perhaps there is a simpler approach?

Anything helps!

Thanks,

John

Answer 1

[Image Analyst]

You can extract the edge pixels with the edge mask like this:

edgePixels = grayImage(edgeMask);

where edgeMask is the binary image (true/false, 1/0, white/black) of your edge locations. Then get the mean value of that by

meanEdgeGrayLevel = mean(edgePixels(:));

[John Moseley]

Thank you very much.

Does extracting the edge pixels produce a list (I am not currently at my MatLab computer)? A list of the gray values for all the edge pixels would be preferable as I am interested in plotting a histogram of the number of pixels (normalized by the total number of edge pixels) (abscissa) vs. gray value (ordinate).

John

[Image Analyst]

It's just a 1D list of all the gray levels under the edge mask. You can just do

[pixelCounts, grayLevels] = imhist(edgePixels, 256);
bar(grayLevels, pixelCounts);
xlim([0 255]);

[John Moseley]

Excellent. Thanks again.

John