Drop an image here or click to browse
PNG, JPEG, WEBP supported
or try a sample →
SVD — Singular Value Decomposition
Each channel is approximated as U·diag(S)·Vᵀ
using only the top k singular values. Higher k → better quality, lower compression.
DCT — Discrete Cosine Transform
Each 8×8 block is DCT-transformed; only the lowest-frequency
keep% coefficients (zig-zag order, like JPEG)
are kept. Lower % → higher compression, more blocking.
Processing…
Compressed Images
Metrics at Current Settings