Image Padding

Updated 02 Oct 2026

•
What it is
- •
  exteds an image beyond its original boundaries using a chosen rule
•
Methods
- •
  zero padding: insert zeros.
- •
  replication: repeat boundary values.
- •
  reflection: mirror values near the boundary.
•
DImensions
- •
  padding every side by $p$ changes an $H\times W$ image to $(H+2p)\times(W+2p)$ .
•
Distinctions
- •
  explicitly adding a border and specifying how a filter handles boundaries are separate operations.
- •
  the project’s padding endpoint creates a larger black-bordered image.
- •
  its filter2D calls do not request zero-border handling; they use the library’s default filtering boundary behavior.
•
Caution
- •
  artificial zeros can introduce dark borders or strong boundary transitions
•
Related
- •
  Spatial Filtering
- •
  Convolution
•

Linked references 3