LLM-Driven Heuristic Frame-Level Quantization Parameter Adaptation for VVenC

2026-06-18

AI Summary

This work addresses the limitations of existing frame-level quantization parameter (QP) allocation in video coding, which struggles to adapt to content characteristics, and the suboptimal performance of traditional Lagrangian rate-distortion optimization due to inaccurate multiplier settings. The authors propose a closed-loop evolutionary framework that leverages large language models (LLMs) to automatically generate and iteratively refine heuristic rules for QP selection, expressed as executable code. These rules are evaluated within the VVenC encoder using statistics from both historical and current frames to provide feedback for evolution. The framework discovers a novel rate-distortion optimization strategy incorporating an entropy term to suppress QP fluctuations, surpassing handcrafted designs. Experiments demonstrate consistent and significant improvements over fixed-QP and conventional baselines across multiple test sets, confirming the effectiveness of LLM-driven algorithmic design in video compression.

Abstract

Optimal frame-level quantization parameter (QP) allocation remains a persistent challenge in modern video encoders. The fixed-QP scheme widely adopted in practical systems is inherently content-agnostic, while classical Lagrangian rate-distortion optimization (RDO) methods often suffer from inaccurate multiplier settings. In this paper, we explore the use of large language models (LLMs) to automatically design RDO heuristics for frame-level QP adaptation. We construct a closed-loop evolutionary framework in which the LLM iteratively proposes RDO heuristics as algorithmic ideas with executable code, and these candidates are evaluated directly through encoding with the Fraunhofer Versatile Video Encoder (VVenC), where each heuristic acts as a scoring function that compares different QP choices based on the encoding statistics of past frames and current candidates. Experimental results across multiple test sets show that the evolved heuristic achieves promising rate-distortion improvements over both the fixed-QP scheme and the Lagrangian baseline. Further analysis reveals that the LLM can autonomously discover an adaptive heuristic that penalizes QP fluctuations via entropy-based terms, providing new insights into the design of RDO algorithms

Problem

• quantization parameter
• rate-distortion optimization
• frame-level QP allocation
• video encoding
• content-adaptive

Innovation

• LLM-driven optimization
• frame-level QP adaptation
• rate-distortion optimization
• evolutionary heuristic search
• VVenC