Canny edge detector

• most widely used in CV for Edge Detection
• theoretical model: optimal when pixels have additive gaussian noise
• the first derivative of the gaussian DoG closely approximates the operator that optimizes the product of signal-to-noise ratio

1. Suppress noise
   1. Filter with x,y derivative of the gaussian DoG
2. Compute Gradient magnitude and direction
3. Apply non-maximum suppression
   1. assures minimal response
   2. Reduce multi-pixel wide edges down to single pixel edge
4. Use hysteresis thresholding and connectivity analysis for Edge Detection

result

• connected edges, thin and non-redundant

• hyperparameter of gaussian filter (kernel) size