Grayscale Image - Border detection \ Filling a specific region with a colour

Dear all,

I have a bunch of grayscale binary images of cross sectional views of an artery (Image 1).

What I am trying to do is to detect the artery borderline, shown in red in image 2 and then digitize the detected borderline in terms of X Y values.

Or

to fill the vessel region with a colour as shown in image 3.

I would be very grateful if anyone would help me with this.

Thanks, Ashkan

Answer 1

[Image Analyst]

Is it going to always be a perfect circle? Or can it be any crazy shape? If any shape, then can you assume it will always fall into a certain gray level range? If so, just threshold and call imfill(), then call bwboundaries().

[Image Analyst]

Ashkan's "Answer" moved here since it's a reply to me rather than a new "Answer" to the original question:

Thanks for your response.

Actually it should not be a perfect circle and can be any shape as shown in the attached image.

I am not sure if I can define a certain gray level rang for all images. How can I find the right threshold to exactly detect the vessel region ...

Are you aware of any other method to detect the vessel border on the images like this? I've seen several published articles which did that using Matlab but they did not disclose the method they used.

Thanks

[Image Analyst]

Ashkan, you can try my interactive thresholding application in my File Exchange: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862. Just set the low and high threshold. Then call

binaryImage = grayImage > lowThreshold & grayImage < highThreshold;
imfill(binaryImage, 'holes');
bwboundaries(binaryImage);

[Ashkan Javadzadegan]

Thanks again.

I downloaded the "Thresholding an image" tutorial. It included three codes:

1- threshold_grayscale_image 2- threshold 3- demo_script

I am beginner to Matlab and am confused with which code I should start with to find the threshold range of my images. Any hint would be much appreciated

[Image Analyst]

Put threshold.m and threshold_grayscale_image.m in your current folder or on your search path. Then follow the sample calls, such as:

grayImage = imread('coins.png');
[lowThreshold, highThreshold, lastThresholdedBand] = threshold(83, 255, grayImage );

[Ashkan Javadzadegan]

I analysed one of my images, attached here.

The colour of the vessel region is not fully black and includes several discrete white regions ...

[Image Analyst]

Oh, so the vessel is the outer bright stuff? I thought you wanted the interior, like where the red spot was. Can't you threshold to get the interior? Then we can call imfill90 like I said.

[Ashkan Javadzadegan]

No the vessel is the interior, and I want to make the inetrior bright. But after using the code the exterior part became bright.

I dont know how to threshold to get the interior ...

[Image Analyst]

Very strange. The thresholding app has two scrollbars, one for the lower threshold and one for the upper threshold. Are you saying that when you shifted them down to select a darker gray region, it still highlighted the bright regions and returned the thresholds of the bright region? I think that's highly doubtful and I will have to look into it. What were the gray levels you chose, and what were the ones that it ended up returning and thresholding the image at?

Answer 2

[Yuan Chen]

a basic idea is to Canny detector to find the edge, then use the hough to find the circle.