0

I've been puzzled with this problem for quite a lot time. I went through a lot of topics and didn't find out exactly what I need. So I started this question.

This is the test image I need to extract features from:

The link is the test image I need to extract features from.

And The final product I want to have is something like:

this link.

I have tried:

  1. Threshold the image and segment out the darker region and tried to use adaptive thresholding, it can segment out the wanted feature (thin lines), but the unmasked edges of the darker region will also be included. And since one of them is connected to one of the lines, I wasn't able to separate them apart (tried watershed segmentation).
  2. I have tried different types of filters by themselves but wasn't able to segment out the feature (because the features are darker than the white area but way brighter than the background, and there are the gradually changed areas which are hard for me to sift out).

Any suggestions would be deeply appreciated.

EBH
  • 10,350
  • 3
  • 34
  • 59
TJ2017
  • 1
  • 2
  • is this somehow physically limited or is your imaging setup just bad? what is this? if the scratches are always vertical you can use morphological operators or directional edge filters to remove other things to some extent. template matching is also an option I guess. – Piglet Jun 16 '17 at 08:52
  • @piglet, Thank you so much for your reply. Yes, the imaging is just bad due to the exposure time limit. The scratches are not always vertical. It can rotate and get angled, which means it could be, in extreme situation, a point. I will try template matching. – TJ2017 Jun 16 '17 at 22:04

0 Answers0