Authors: Surafel Wolde (FAN: 8047836790317269) & Aziza Jemal

Academic Target: IEEE Transactions on Multimedia / IEEE Data Compression Conference (DCC)

The Combinatorial Multi-Node Tree (CMNT) is an end-to-end, high-performance image compression architecture. Originally conceptualized from a physical multi-channel sparse transmission model, CMNT bridges closed-loop spatial-color prediction with a non-binary $M$-ary ($M=3$) combinatorial prefix tree and dense base-3 physical integer packing.

On continuous-tone photography, CGI art, and 3D renders, CMNT consistently outperforms standard Deflate/PNG by up to 29.7% in file size with 100% bit-perfect mathematical reversibility.

Zero Third-Party Compression Wrappers: Operates from absolute first principles. No zlib, no gzip, no zipfile. Reads and writes its own native binary .cmnt container.

100% Bit-Perfect Reversibility: Verified via cryptographically identical SHA-256 pixel hashes (assert original == decoded).

Dense Base-3 Physical Byte Packing: Solves the non-binary storage bottleneck by packing 5 ternary trits into a single 8-bit byte with 99.06% theoretical storage density.

Dual-Mode Engine:

Lossless Master Mode: 100% pixel preservation, outperforming standard PNG across continuous-tone graphics.

Compact Web Mode: Closed-loop DPCM near-lossless quantization ($\Delta > 1$) matching/beating JPEG file sizes with zero $8 \times 8$ grid blocking artifacts.

Portable Desktop Studio: Includes CMNT_Studio.exe, a standalone GUI with full-screen viewing, multi-format conversion (PNG, JPG, WebP, BMP, TIFF), and Windows File Explorer registration.

📊 Comprehensive Lossless Benchmarks

Evaluated against standard PNG (Deflate) and Google WebP Lossless on real-world datasets:

Result: With Meta-Adaptive Near-zero Integer Arithmetic Coding (MANIAC) context modeling, the CMNT pipeline achieves a clean sweep over standard PNG across all benchmark images.

[Raw 24-bit RGB Image] │ ▼ ┌────────────────────────────────────────────────────────┐ │ STAGE 1: Reversible Color Space Decorrelation (RCT) │ │ • Y = G, U = R - G, V = B - G │ └────────────────────────────────────────────────────────┘ │ ▼ ┌────────────────────────────────────────────────────────┐ │ STAGE 2: Closed-Loop 2D Adaptive Spatial Prediction │ │ • Median Adaptive Predictor (MAP): │ │ if c >= max(a,b) -> min(a,b) │ │ if c <= min(a,b) -> max(a,b) │ │ else -> a + b - c │ │ • Residual Error: ε = x - x̂ │ └────────────────────────────────────────────────────────┘ │ ▼ ┌────────────────────────────────────────────────────────┐ │ STAGE 3: Combinatorial Multi-Node Tree (CMNT) │ │ • Non-binary 3-ary (M=3) Prefix Tree │ │ • Maps residuals to uniquely decodable prefix trits │ └────────────────────────────────────────────────────────┘ │ ▼ ┌────────────────────────────────────────────────────────┐ │ STAGE 4: Base-3 Dense Trit Physical Packing │ │ • Packs groups of 5 trits into a single 8-bit byte: │ │ Byte = t₀ + 3·t₁ + 9·t₂ + 27·t₃ + 81·t₄ ≤ 242 │ │ • Physical Storage Packing Density: 99.06% │ └────────────────────────────────────────────────────────┘ │ ▼ ┌────────────────────────────────────────────────────────┐ │ STAGE 5: Native Binary Container Serialization (.cmnt) │ │ • Magic signature: b'CMNT' │ └────────────────────────────────────────────────────────┘

🛠️ Installation & Quick Start

1. Requirements

git clone https://github.com/your-username/cmnt-codec.git cd cmnt-codec pip install -r requirements.txt

2. Command Line Interface (CLI)

Compress an image to .cmnt (100% Bit-Perfect Lossless)

python -m cmnt.cli compress input.png output.cmnt

Compress in high-efficiency compact lossy mode

python -m cmnt.cli compress input.png output.cmnt --lossy

Decompress .cmnt back to PNG

python -m cmnt.cli decompress output.cmnt restored.png

View metadata of a .cmnt file

python -m cmnt.cli info output.cmnt

Open directly in Windows Image Viewer

python -m cmnt.cli view output.cmnt

3. Launch Desktop GUI Studio

python -m cmnt.gui

Or run the standalone Windows binary: CMNT_Studio.exe.

To run the automated bit-perfect verification suite:

Bit-perfect match: True Original SHA-256: adbf468d64ac2729b7c5dfe8a93d9ac491cf6f063c011c2b786a6fd55ca73130 Restored SHA-256: adbf468d64ac2729b7c5dfe8a93d9ac491cf6f063c011c2b786a6fd55ca73130

SUCCESS: 100% BIT-PERFECT RECONSTRUCTION FROM REAL .CMNT FILE! <<<

📄 Academic Citation

If you use or reference this architecture in your research, please cite:

@article{wolde2026cmnt, title = {Joint Combinatorial Partition Signaling for Next-Generation Lossless Image and Video Compression}, author = {Wolde, Surafel and Jemal, Aziza}, year = {2026}, note = {Available at GitHub: https://github.com/your-username/cmnt-codec} }

⚖️ License

Distributed under the MIT License. See LICENSE for details.