The Standard Hough Transform
- Common method for extraction of geometric primitives
- Voting-based algorithm
- Each active input pixel is mapped to accumulator space
- Accumulator space has dimension = number of parameters of structure
- Circle: (x - x1)2 + (y-y1)2 = r2 Has three parameters: x1, y1, r.
For fixed radius circle detection, a single pixel is mapped to a circle in the accumulator (1)
- For multiple radii circle detection, input pixels are mapped to cones
- Each bin in the accumulator represents a circle on which the the input pixel is potentially a point.
SHT algorithm votes all of active input pixels into accumulator
- Inverse transform can then be run on accumulator to reconstruct circles
Figure #1: The mapping of a single active pixel into accumulator space is a circle.