1. Detection of Isolated Points
Purpose:
To detect pixels that have intensity significantly different from their 8 neighbors — these are isolated points.
Method Used:
Use second-order derivative — specifically, the Laplacian Operator:
2. Line Detection
Purpose:
Detect thin, one-pixel wide lines in a specific direction: horizontal, vertical, or diagonal (±45°).
Method Used:
Use directional second derivative kernels (3×3 masks). Four common kernels are:
3. Edge Detection
Purpose:
Detect boundaries between regions with different intensity levels.
Edge Models:
Edges are classified by intensity profile:
(a) Step Edge:
Sudden jump in intensity.
Occurs over 1 pixel (ideal edge).
(b) Ramp Edge:
Gradual intensity change (blurred step).
Occurs over multiple pixels.
(c) Roof Edge:
Thin bright/dark line with a peak (often 1-pixel wide).
Derivative-Based Detection:
- First Derivative (Gradient):
- Magnitude is high at edges.
- Used for edge detection.
- Produces thick edges.
- Second Derivative (Laplacian):
- Zero crossing corresponds to edge center.
- Produces thin, sharper edges.
- Used in Laplacian or Canny edge detectors.