Line Detection

• A lot of work goes into line and edge detection.
• When I was in graduate school (in the 90s), I used to listen to a different PhD student talking about their method each week.
• These are typically worried about straight lines.
• There is a separate area for working with curves.
• Images are not perfect, so there are often a lot of bumps in the image (which our eyes cancel out subconsciously).
• Averaging (see blending) is a useful technique to cope with noise.
• For an edge, you want to see a permanent shift.
• For a line, you want to see a one (or a few) off shift.
• You typically have to cope with the full range of angles, and you want to cover different areas.
• There is also a field of connecting partial lines.
• As usual, the more you know about the image, the easier it is to work with.