Serialization and compression

JSON serialization of trained bases (byte-compatible with the Julia package) and top-k lossy compression of images in a learned basis.

Serialization (JSON) and lossy compression of trained bases.

Contents

save_basis	Write basis to a JSON file.
load_basis	Read basis from a JSON file.
basis_to_dict	Serialize a QFTBasis to a plain Python dict (JSON-ready).
dict_to_basis	Inverse of basis_to_dict.
basis_hash	SHA256 of a canonical string form, matching Julia's basis_hash(QFTBasis).
format_float_julia_like	Format a Python float the way Julia's string(x::Float64) does.
compress	Compress image under basis, keeping top (1 - ratio) fraction.
compress_with_k	Compress keeping exactly k coefficients.
recover	Reconstruct real-valued image from a compressed representation.
CompressedImage	Sparse frequency-domain representation of an image.
compression_stats	Summary of compression ratio + kept coefficients.
save_compressed	Write a CompressedImage to path as JSON; returns the path.
load_compressed	Read a CompressedImage written by save_compressed.
compressed_to_dict	Convert to the JSON-ready dict form (schema version _VERSION).
dict_to_compressed	Inverse of compressed_to_dict.