# Enhancing very thin lines in images

I want to detect filopodia (red arrow indicates filopodia in a vessel image).

My strategy is that if $I$ is the original and $J$ is the another image after opening operation. Then, $I-J$ is the filopodia. (See Figure 6 in this paper).

To do this, I want to enhance the very thin lines. How to do it? I also have information of front line like this:

Or, do you have any idea for detecting filopodia?

• define "very smalll widths" in pixels; if they are in the order of single pixels, you will have a hard time based on frequency content alone. – Marcus Müller Aug 15 '16 at 12:05
• sorry for ambiguity. It means "very thin". – jakeoung Aug 15 '16 at 14:39
• what's "very thin" in pixels? (the wording is totally OK, just: numbers are better than words like "small" or "thin") – Marcus Müller Aug 15 '16 at 16:27
• there is no exact definition for that, because the terminology is from the context of biology. Red arrows indicate filopodia which I want to segment. – jakeoung Aug 15 '16 at 16:53
• the point is that this is a DSP problem, so a bit of info on how the problem behaves digitally is necessary. – Marcus Müller Aug 15 '16 at 16:58