Is there a standard method for finding long thin objects in a noisy
image? I need to measure the length in many images automatically. I can
see them by eye easily, and can sometimes even get a good binary image
after a lot of work. I have had some preliminary success with the
following algorithm: at each point, take segments of length 2K+1 pixels,
at different angles, so the segments end centered in a pixel. There are
4k of these. Take the average pixel value along each segment (using linear
interpolation where the segment doesn't land on a pixel). This gives
4k numbers. Choose the smallest and use that value as the output value for
the center point. The resulting image enhances long thin objects, although
it also sometimes creates thin objects where there weren't any, for example,
it will connect two bright spots with a phantom line.
Is there a standard way of solving this problem?
