In this lesson, we focus the use of three important concepts used for image segmentation: local minimum, skeleton cropping and image reconstruction.
The goal in this experiment is to identify and extract the thin horizontal lines from the original gray-scale image. The approach used in this lesson is to first identify local minima in the vertical direction, as the lines of interest are horizontal and dark in a lighter background.
The local minima image is very noise and we can eliminate the noisy lines by doing an image reconstruction using as marker a homotopic skeleton crop operator. In this way, we eliminate smaller connected lines. An alternative way of finding the marker would be to apply an area opening, which eliminates the connected lines (regions) smaller than a specified area.
We show below the original gray-scale image and the result of the application of a local minimum detection using a vertical 1x3 structuring element.
|
|
|---|---|
| a) | b) |
We crop the skeletons by using the template to detect end-points:
X X X 0 1 0 0 0 0
We use the cropped skeleton as the marker for the binary image reconstruction.
|
|
|---|---|
| a) | b) |
We can see that this method does not give the best performance in the segmentation process as the image is very noise and there are lines which runs in the diagonal direction, specially in the borders of the image.