This is an open-source implementation of the Line Segment Detector paper by Grompone von Gioi et al. This is a complete reimplementation from scratch, as the original source code publsihed with the paper is under the Aferro GPL license, which is much too restrictive for common use cases. As a result, this project is published under a much more permissive MIT license.

Pending release on PyPi, the library can be installed directly from GitHub:

pip3 install git+https://github.com/hdkai/Line-Segment-Detector.git

The library exposes a single method:

def line_segment_detector (image: ndarray, scale=0.8, angle_tolerance=22.5):
    """
    Compute line segments in an image using a modified implementation of the 
    Line Segment Detector paper: http://www.ipol.im/pub/art/2012/gjmr-lsd/article.pdf

    Parameters:
        image (ndarray): Input image, either RGB or greyscale.
        scale (float): Downscale factor for processing.
        angle_tolerance (float): Angle tolerance in degrees.

    Returns:
        ndarray: Lines array with shape (N,8), with each row being [ x1, y1, x2, y2, cx, cy, l, w ].
    """

The function accepts a single color or greyscale image, and returns an Nx8 matrix, where each row contains:

1. The x coordinate of the first point on the line segment.
2. The y coordinate of the first point on the line segment.
3. The x coordinate of the second point on the line segment.
4. The y coordinate of the second point on the line segment.
5. The x coordinate of the center of the line segment.
6. The y coordinate of the center of the line segment.
7. The length of the line segment l.
8. The width of the line segment w. For every detected line, w < l.

• Python 3.6+
• Cython 0.29.21+.