Highlighting Scratches/Digs in an Image

[Steve.Briggs]

Hey everyone,

I'm taking images using an ISG_Corp LW1.3-SM Camera and acquiring images through a prism that reflects light back through into an mms.  The image I get back shows a variety of scratches and digs in the prism (cuts and dots) which indicate imperfections in the coating on the prism.  Attached is an image that has a number of black and white dots/circles that are digs.

What would be the easiest way to run a vi that would be able to highlight a number of these imperfections that are present in the image.  Any help is greatly appreciated.

[BruceAmmons]

This is a good situation to use local adaptive thresholding.  I would run it once looking for dark objects, and again looking for bright objects.  This will do a good job of highlighting all your defects, I believe.

Bruce

[vista]

Hello  Have a nice day!

[Steve.Briggs]

After a lot of investigation and altering of my image acquisition and illumination methods, I've made a great deal of progress.  I switched to a backlight illumination technique so the imperfections in my prism now appear white, and everything else is black/dark grey.  I also took a lot of what you provided Vista and worked onward with it along with examples from Blob Analysis.

What I'm curious about now is a way of making my image processing steps easier...when I go through edge detection or try to enhance the edges in my image, a lot of unwanted edges are detected because my "black" regions aren't completely black.  They have a pixel depth ranging from 25-35, whereas my imperfections, the areas I am actually interested in, have a pixel depth ranging from 140 - 255, and I'm able to handle the saturation.

I need to be able to somehow decrease the pixel/image depth of the "black" areas with values between 25-35 to a value at or around 0, so my edge enhancement isn't affected by unwanted material.

Any help or suggestions are greatly appreciated.

Attached is a current image...

[BruceAmmons]

So subtract 100 from your image.  All the black areas will go to zero, and your white range will be 40 - 155.

Once again, this would be an ideal application for localized thresholding.  Possibly even simple auto thresholding.

Bruce

[Steve.Briggs]

---

@Bruce Ammons wrote:

So subtract 100 from your image. 

---

Shows how much I know when it comes to vision and imaging.  Thanks Bruce, had no idea it was that easy.

I do utilize thresholding, but like I said, the darkened regions that I do not want come into play.  Perhaps I do it in the wrong fashion, I'll explore it more once I adjust the image and subtract the 100 from it.