3 filtering facts that show up in exams

Lesson notes

Key points

• 3. Mean filter blurs: A 3×3 averaging kernel smooths noise, but it also blurs edges because it averages across them.
• 2. Median beats salt-pepper: The median filter removes salt-and-pepper noise best, because extreme outlier pixels never become the median.
• 1. Edge kernels sum to zero: Sharpening and edge kernels like Laplacian sum to zero, so flat regions output zero and only edges remain.

Takeaway: Median for impulse noise; zero-sum kernels detect edges.