PISE: Physics-Anchored Semantically-Enhanced Deep Computational Ghost Imaging for Robust Low-Bandwidth Machine Perception

2026-01-18

AI Summary

This work addresses the significant degradation in ghost imaging quality and downstream classification performance caused by extremely low sampling rates on bandwidth-constrained edge devices. To overcome this challenge, the authors propose a deep ghost imaging method that synergistically integrates physical priors with semantic guidance—marking the first effort to jointly embed physical information and semantic supervision into a computational ghost imaging framework. The approach constructs a physics-constrained network via adjoint operator initialization and introduces a semantic-enhanced deep reconstruction architecture, enabling high-quality image recovery at merely 5% sampling rate. Experimental results demonstrate a 2.57% improvement in classification accuracy and a ninefold reduction in result variance compared to existing methods, substantially enhancing imaging stability and perceptual robustness.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

We propose PISE, a physics-informed deep ghost imaging framework for low-bandwidth edge perception. By combining adjoint operator initialization with semantic guidance, PISE improves classification accuracy by 2.57% and reduces variance by 9x at 5% sampling.

Problem

• ghost imaging
• low-bandwidth perception
• edge computing
• machine perception
• computational imaging

Innovation

• physics-informed
• computational ghost imaging
• semantic guidance
• low-bandwidth perception
• adjoint operator initialization