README.md

Complexity Ratio Metric for VVC Encoder Evaluation

This repository provides the implementation of the Complexity Ratio Metric $C_R$ proposed in
M. E. A. Kherchouche, F. Galpin, T. Dumas, F. Schnitzler, L. Zhang, and D. Menard,
"Neural Network-based Rate-Distortion Costs Approximation for Intra QTMTT Partitioning,"
IEEE Transactions on Circuits and Systems for Video Technology, 2025.

---

Overview

The Complexity Ratio ($C_R$) is a hardware-independent metric designed to quantify encoder complexity reduction between a reference encoder (e.g., VTM) and an accelerated version.
Unlike encoding time, which depends on the execution environment, $C_R$ measures only the algorithmic workload within the Rate–Distortion Optimization (RDO) process.

The metric compares the cumulated number of reconstructed pixels processed through RDO between two encoder variants, providing an objective and reproducible measure of complexity reduction.

---

Formula

$$ C_R = 100 - \frac{C_a -C_p}{C_a} $$

where:

• $C_p$: Cumulated number of reconstructed pixels through the RDO of the encoder + the acceleration method
• $C_a$: Cumulated number of reconstructed pixels through the RDO of the default encoder (e.g., VTM)

---

Citation

If you use this metric or code in your research, please cite the following paper:

@techreport{ComplexityMetricRatio,
  author    = {M. E. A. Kherchouche and F. Galpin and T. Dumas and L. Zhang and D. Menard},
  title     = {Complexity Ratio for VVC},
  institution = "InterDigital",
  year      = {2025},
  url       = {https://github.com/InterDigitalInc/ComplexityRatio}
}