Combining geometric and photometric information to find lines from step edge detection

This paper deals with the problem of detecting both step and line edges using a classical step edge detector. As a gradient detector produces two extrema when applied to a line, we propose a method which is based on such a detector but which appropriately responds to both step and line edges. The idea is first to identify line contours using geometric and photometric properties and second to substitute a line for a single contour using a skeletonization algorithm.<>