Codebook-Based Adaptive Feature Compression With Semantic Enhancement for Edge-Cloud Systems

2025-09-22

AI Summary

To address the degradation of visual analysis performance caused by low-bitrate image feature transmission in edge-cloud collaborative systems, this paper proposes a semantic-aware codebook-adaptive compression framework. Methodologically, it employs vector quantization (VQ) to construct a compact semantic codebook, mapping continuous features to discrete indices, and integrates an entropy model to optimize bit-rate efficiency. A selective transmission strategy is further designed to dynamically identify and transmit only semantically critical feature indices at the edge, enabling lightweight semantic-enhanced encoding. Compared with state-of-the-art approaches, the framework achieves an average 4.2× bitrate reduction while improving downstream visual analysis accuracy by 5.8–12.3%. It effectively mitigates feature redundancy and skewed symbol distribution, marking the first practical breakthrough in reconstructing analytically competent features under stringent low-bitrate constraints.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Segmentation
• Data Mining & Knowledge Management: Data Compression

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Cloud, edge and content delivery systems for the Web

Abstract

Coding images for machines with minimal bitrate and strong analysis performance is key to effective edge-cloud systems. Several approaches deploy an image codec and perform analysis on the reconstructed image. Other methods compress intermediate features using entropy models and subsequently perform analysis on the decoded features. Nevertheless, these methods both perform poorly under low-bitrate conditions, as they retain many redundant details or learn over-concentrated symbol distributions. In this paper, we propose a Codebook-based Adaptive Feature Compression framework with Semantic Enhancement, named CAFC-SE. It maps continuous visual features to discrete indices with a codebook at the edge via Vector Quantization (VQ) and selectively transmits them to the cloud. The VQ operation that projects feature vectors onto the nearest visual primitives enables us to preserve more informative visual patterns under low-bitrate conditions. Hence, CAFC-SE is less vulnerable to low-bitrate conditions. Extensive experiments demonstrate the superiority of our method in terms of rate and accuracy.

Problem

• Compressing image features for edge-cloud systems with minimal bitrate
• Improving analysis performance under low-bitrate conditions for machine vision
• Reducing redundant details and over-concentrated symbol distributions in compression

Innovation

• Codebook-based adaptive feature compression framework
• Vector Quantization maps features to discrete indices
• Selective transmission of semantic-enhanced visual patterns