This repository contains scripts to calculate the famous Bjontegaard-Delta metric with different interpolation functions.
-
Updated
Jul 3, 2024 - Python
This repository contains scripts to calculate the famous Bjontegaard-Delta metric with different interpolation functions.
Matlab code to compute entropy, conditional entropy, mutual information, channel capacity, and rate distortion.
Reproducible research toolkit for benchmarking AVIF, JPEG XL, WebP, and JPEG on web images with perceptual quality metrics, statistical analysis, visual comparisons, and interactive reports.
Neural Distributed Compressor Discovers Binning [JSAIT 2024]
Implementations of different ways to calculate Shannon's rate-distortion trade-off
A source-aware rate-distortion index of image and video tokenizers.
Observation Theory — registration-first claim ledger, prediction registry, experiments, and the foundational paper (Book XIV of the Geometric Series).
MATLAB image compression: AEV-DLT v3 per-coefficient fair-rate PCRD (CDF 9/7 wavelet) - beats EBCOT Tier-2 baseline on boat & peppers
Compute useful internal structure (C_u) for thermodynamic intelligence. Exact IB-optimal toy + neural network validation.
Observation Theory — the front door: thesis, method, and the map of the ledgered repositories. The theory has one public name; every claim resolves to a ledger row.
Code for signal recovery from quantized sparsified measurements with imperfect side information
Rigorous bounds on the gap between learned neural codec rate and the optimal rate-distortion function R(D).
The following project is an analytical analysis which compares deep learned image compression (CompressAI, Conv2D Autoencoder) against traditional codecs (JPEG, PNG) on CIFAR-10 on the basis of the benchmarks: rate-distortion and latency.
Three MQTT data-transmission systems using learned and classical compression — latent-space image codec, Relative Entropy Coding, and REC over a VAE latent — each with sender/receiver apps, full analysis, and documentation.
Honest RD evaluation for compression research: gate-checked Bjontegaard (BD-rate) metrics, matched-quality single-point comparison, publication-style RD plots
Information-theoretic lower bounds for diffusion-based image compression with learned generative priors.
A from-scratch deep learning image compression prototype using a convolutional autoencoder, latent quantization, entropy modeling, and Huffman coding.
Extends high-rate asymptotic theory of Nonlinear Transform Coding (NTC) to finite-rate regimes with numerically verified bounds.
Projection as Conditional Expectation and the Capacity Ceiling (Cosmochrony non-injective foundations)
PyTorch-based neural video compression prototype built from scratch with motion estimation, residual coding, entropy modeling, and rate-distortion optimization.
To associate your repository with the rate-distortion topic, visit your repo's landing page and select "manage topics."