Hough Transform for Edge Detection

GIF source: 16720B course materials, Prof. Kris Kitani (CMU)

This repo contains a from-scratch implementation of hough transform. (Implemented as part of the Computer Vision course 16720B at CMU)

Concept:

If two edge points lay on the same line, their corresponding cosine curves will intersect each other on a specific $(ρ, θ)$ pair. Thus, the Hough Transform algorithm detects lines by finding the $(ρ, θ)$ pairs that have a number of intersections larger than a certain threshold.

Repo structure:

1. Run detect_edges.py for Edge detection with Non-Maximal Suppression(NMS) (NMS is done to get sharper edges)

2. Run detect_houghlines.py for applying Hough transform with Non-Maximal Supression(to supress neighbouring hough lines) on the edge magnitude image to identify the hough lines

3. Run visualize_ls.py to visualize intersection of hough lines and edges

Sample output:

          Original image:

          

1. Edge detection with NMS

2. Detected Hough lines using Hough Transform

3. Interesection of Hough lines with edges

Few more final results: