Measuring line length using ridge detection

ridge-detection
binary
skeleton-analysis
thresholding
Tags: #<Tag:0x00007fd69cacb9d8> #<Tag:0x00007fd69cacb898> #<Tag:0x00007fd69cacb758> #<Tag:0x00007fd69cacb550>

#1

Hello,

I am trying to measure the line length of the blue lines shown in the image below. I have tried using various combinations of making it binary, auto threshold, using skeleton 3d, and other image altering processes to get ridge detection to properly read the lines, but I am having no luck. I always end up with ridge detection saying I have more blue lines than I actually do. Has anyone had any luck with a similar project that can point me in the right direction or is using a different method to alter the picture so ridge detection can properly work?

Any help is greatly appreciated!


#2

Good day!

I guess your image doesn’t show real data …

What precision do you expect?

I would use the perimeter and an estimate of the width of the structures. With these measures the length could be estimated as
(perimeter - 2 * width) / 2

Regards

Herbie


#3

Hello Herbie,

I have attached a link to a picture of what i want the image process to work on. I was using the above picture since it as a lot easier to validate whether the length is correct or not. Please let me know whether your advice would change having viewed this photo. I am looking for it to be decently precise, anything above 80% would be great.

http://forum.imagej.net/uploads/default/original/2X/b/b09c6f02d4173c1ebac75e387fd15d50c002468a.jpg

Thanks!


#4

Funny,

we had this task before:
http://forum.imagej.net/t/great-overlaid-rigid-rod-segmentation-challenge/3829/19?u=herbie

Good luck

Herbie


#5

It seems to work well on that picture, but i cannot get it to work on pictures such as below


#6

I guess the question is whether there is a specific method that would be possible, or if i need to look into changing the pictures to provide a larger contrast between the lines and the background since i haven’t been able to do so with threshold


#7

Good day!

the fundamental problem with this image is noise and spatial resolution. The former is much too high and the latter is rather low …

Best

Herbie


#8

The provided link to an earlier forum discussion
http://forum.imagej.net/t/great-overlaid-rigid-rod-segmentation-challenge/3829
contains a number of approaches that you should consider.

In my opinion the main problem with images such as the one you refer to here:
http://forum.imagej.net/uploads/default/original/2X/b/b09c6f02d4173c1ebac75e387fd15d50c002468a.jpg
are overlapping/crossing/joining fibers (and perhaps low spatial resolution). Both problems are inexistent with your introductory task.

If you study the proposed approaches you will realize that a solution of the task is far from trivial.

Regards

Herbie