Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

How2Compress: Scalable and Efficient Edge Video Analytics via Adaptive Granular Video Compression

Yuheng Wu, Thanh-Tung Nguyen, Lucas Liebe, Quang Tau, Pablo Espinosa Campos, Jinghan Cheng, Dongman Lee

TL;DR

How2Compress advances edge video analytics by introducing fine-grained, macroblock-level quality control that is self-supervised and region-aware. By reframing emphasis assignment as a segmentation task and coupling Region-aware Emphasis Routing with a Proxy Emphasis Target, it achieves substantial bitrate reductions (up to $50.4\%$) while preserving task accuracy across detection, tracking, and pose estimation. The approach is designed to be plug-and-play with widely deployed codecs like H.264, with robust performance across diverse scenes and hardware, and shows favorable latency and computational cost for real-time edge deployment. This work offers a practical pathway to scalable, machine-centric video compression in edge computing environments, with generalization to other codecs and downstream tasks and a clear structure for future refinements.

Abstract

With the rapid proliferation of the Internet of Things, video analytics has become a cornerstone application in wireless multimedia sensor networks. To support such applications under bandwidth constraints, learning-based adaptive quantization for video compression have demonstrated strong potential in reducing bitrate while maintaining analytical accuracy. However, existing frameworks often fail to fully exploit the fine-grained quality control enabled by modern blockbased video codecs, leaving significant compression efficiency untapped. In this paper, we present How2Compress, a simple yet effective framework designed to enhance video compression efficiency through precise, fine-grained quality control at the macroblock level. How2Compress is a plug-and-play module and can be seamlessly integrated into any existing edge video analytics pipelines. We implement How2Compress on the H.264 codec and evaluate its performance across diverse real-world scenarios. Experimental results show that How2Compress achieves up to $50.4\%$ bitrate savings and outperforms baselines by up to $3.01\times$ without compromising accuracy, demonstrating its practical effectiveness and efficiency. Code is available at https://github.com/wyhallenwu/how2compress and a reproducible docker image at https://hub.docker.com/r/wuyuheng/how2compress.

How2Compress: Scalable and Efficient Edge Video Analytics via Adaptive Granular Video Compression

TL;DR

How2Compress advances edge video analytics by introducing fine-grained, macroblock-level quality control that is self-supervised and region-aware. By reframing emphasis assignment as a segmentation task and coupling Region-aware Emphasis Routing with a Proxy Emphasis Target, it achieves substantial bitrate reductions (up to ) while preserving task accuracy across detection, tracking, and pose estimation. The approach is designed to be plug-and-play with widely deployed codecs like H.264, with robust performance across diverse scenes and hardware, and shows favorable latency and computational cost for real-time edge deployment. This work offers a practical pathway to scalable, machine-centric video compression in edge computing environments, with generalization to other codecs and downstream tasks and a clear structure for future refinements.

Abstract

With the rapid proliferation of the Internet of Things, video analytics has become a cornerstone application in wireless multimedia sensor networks. To support such applications under bandwidth constraints, learning-based adaptive quantization for video compression have demonstrated strong potential in reducing bitrate while maintaining analytical accuracy. However, existing frameworks often fail to fully exploit the fine-grained quality control enabled by modern blockbased video codecs, leaving significant compression efficiency untapped. In this paper, we present How2Compress, a simple yet effective framework designed to enhance video compression efficiency through precise, fine-grained quality control at the macroblock level. How2Compress is a plug-and-play module and can be seamlessly integrated into any existing edge video analytics pipelines. We implement How2Compress on the H.264 codec and evaluate its performance across diverse real-world scenarios. Experimental results show that How2Compress achieves up to bitrate savings and outperforms baselines by up to without compromising accuracy, demonstrating its practical effectiveness and efficiency. Code is available at https://github.com/wyhallenwu/how2compress and a reproducible docker image at https://hub.docker.com/r/wuyuheng/how2compress.
Paper Structure (47 sections, 4 equations, 15 figures, 10 tables, 2 algorithms)

This paper contains 47 sections, 4 equations, 15 figures, 10 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background
  3. Preliminaries on Video Compression
  4. Related Work
  5. Measurement Study
  6. Problem Formulation
  7. Methodology
  8. Emphasis Assignment Model (EA Model)
  9. Region-aware Emphasis Routing (RER)
  10. Proxy Emphasis Threshold (PET)
  11. Region-aware Dual Exploration
  12. Evaluation
  13. Experimental Setup
  14. Evaluation Results
  15. Better Bitrate Savings across Diverse Scenes
  16. ...and 32 more sections

Figures (15)

  • Figure 1: Machine-centric video compression is a key component to unleash wireless sensor network efficiency.
  • Figure 1: Comparison between our proposed framework and prior works.
  • Figure 3: Samples from MOT dataset (a-d). Bitrate savings over raw video using human-centric fine-grained (e) and machine-centric coarse-grained (f) compression.
  • Figure 4: Overview of How2Compress.
  • Figure 5: Design motivation: (a) Different macroblocks ex- hibit varying resilience to compression based on content characteristics. (b) Macroblocks in RoI are more sensitive.
  • ...and 10 more figures