Hybrid Images

A classic computer-vision build: hybrid images combine the low-frequency components of one image with the high-frequency components of another, so the picture you see changes with viewing distance: one face up close, a different one from across the room.

How it works

1. Build Gaussian and Laplacian filters from scratch (no black-box calls).
2. Low-pass one image, high-pass the other.
3. Align, weight, and sum the frequency bands into a single hybrid.

The interesting engineering is in the details: kernel size vs. cutoff frequency trade-offs, alignment quality, and making the pipeline reusable across image pairs.