kernel

mathematics signal_processing

A kernel is a small matrix used in convolution operations. Each element is a weight that determines how much a neighboring pixel contributes to the output. A 3×3 Gaussian kernel smooths by weighting the center pixel most. A Sobel kernel detects edges by computing intensity gradients. The choice of kernel defines the operation — the convolution mechanism is identical regardless.

Examples

  • 3×3 Gaussian kernel for blur
  • Sobel kernel for edge detection
  • Laplacian kernel for sharpening
  • Identity kernel (no change)

Appears In

Related Terms