Extract feature area from image

Question

I have an image like the one above, and what I need to find are the red circled areas. They are caused by how the nerve layer is in the eye, and I have no idea what method to use to extract them. (I am looking for a brighter background from the nerve layer) I have found all the blood vessels, so I can ignore those area easily. But I don't have any concept of how to begin the search for these areas. Note that the bright spots on the right top side (lets say 12 o clock / 1 o clock from the image center) are not caused by the nerve layer, but from being in a young persons eye.

Can someone point me to a method to extract these areas? Note that the areas could also be non-existant, or be formed like a cross, or like an hourglass.

Answer 1

It is going to be a tough task, and you may not achieve an automated process.

Here are some bricks that you could try:

1. Enhance the contrast of the image so that you can clearly distinguish all the features by looking at the image. Here is a link to an algorithm that shows good results.

2. In this enhanced image, extract the edges (using for example a standard Canny-Deriche edge detector).

3. From the edge detector output, you can start adding the required application logic in order to keep the desired vessels: size or length criteria, linear features extraction, etc.