Image Compression
via SVD

Upload any image and watch Singular Value Decomposition strip away redundancy in real time. Adjust rank k to find the sweet spot between file size and visual quality.

A ≈ U_k · Σ_k · V^T_k rank-k approximation

1Upload any PNG, JPG or WEBP image

›

2Browser computes Jacobi SVD per channel

›

3Drag slider to choose rank-k approximation

›

4View metrics, error map & download lossless PNG

⬆

Drop image here

PNG, JPG, WEBP — up to 2048×2048
Or click to browse files

Original —

Compressed (rank k = —) —

Error map (|original − compressed| × 5) difference

Computing SVD...

Rank k

k 20

Channel mode

Metrics

Compression

—

PSNR (dB)

—

Energy %

—

Saved

—

Singular value spectrum

PSNR vs k (quality curve)

Display

Show error map

Export

Image Compressionvia SVD

Drop image here

Image Compression
via SVD