BLUE: Semantics-Preserving Video Compression for Efficient Vision-Language Surveillance Analytics

2026-07-21

AI Summary

This work addresses the limitations of traditional video compression, which often degrades semantic understanding by downstream vision-language models (VLMs) and fails to reduce inference calls. To overcome these issues, the authors propose BLUE—the first semantics-preserving compression method tailored for machine vision in fixed-camera surveillance scenarios. BLUE leverages foreground-background separation to suppress redundancy from static backgrounds while preserving semantically critical activity content. Built upon the H.265 framework, it integrates content-aware compression with a packet-size-driven frame-skipping strategy. Experiments on the VIRAT and CHAD datasets demonstrate that BLUE reduces VLM inference calls by an estimated 53% without significant loss in semantic quality, thereby achieving decoupled optimization of compression efficiency and semantic fidelity.

Abstract

Continuous surveillance video creates a growing storage, transmission, and inference burden for enterprise video analytics systems. While modern codecs such as H.265 reduce bitrate for human-viewable video, aggressive compression can degrade downstream computer-vision performance and does not necessarily reduce the number of vision-language model (VLM) inference calls required for semantic video understanding. This paper evaluates BLUE, a fixed-camera surveillance compression approach that suppresses static-background redundancy while preserving foreground activity, for its effect on VLM-based event and anomaly understanding. We compare raw H.265 and BLUE-compressed H.265 video on two surveillance datasets: VIRAT, comprising 227 paired event samples from 106 clips, and CHAD, comprising 54 human-activity anomaly clips. For each pair, the same frame index is evaluated using a VLM captioning pipeline, and outputs are scored against annotation-derived ground truth using a blind judging protocol. The results show no measurable degradation in semantic inference quality. On VIRAT, the mean VLM score remains effectively unchanged between raw H.265 and BLUE, with a mean difference of approximately -0.01 on a 0-10 scale. On CHAD, raw H.265 and BLUE obtain near-equivalent mean scores of 4.31 and 4.26, respectively. Compression saving is also uncorrelated with VLM score change on VIRAT (r = 0.004), indicating that higher BLUE compression does not predict semantic quality loss. Beyond storage reduction, BLUE increases the share of skip-heavy P-frames on CHAD from 1.4% to 53.2%, enabling an estimated 53% reduction in VLM calls through packet-size-based frame skipping. These findings suggest that BLUE functions as a machine-centric compression layer for surveillance video, reducing bandwidth and inference cost while preserving VLM semantic performance.

Problem

• video compression
• vision-language models
• surveillance analytics
• semantic preservation
• inference efficiency

Innovation

• semantics-preserving compression
• vision-language models
• surveillance video analytics
• machine-centric compression
• frame skipping