PS-EIP: Robust Photometric Stereo Based on Event Interval Profile

2025-03-24

AI Summary

This work addresses the sensitivity of event-camera photometric stereo to noise, cast shadows, and non-Lambertian reflections—arising from independent processing of inter-event intervals. We propose an unsupervised, non-deep-learning method that explicitly models pixel-wise temporal event-interval profiles for the first time. Our core contributions include: (i) a contour continuity constraint enforcing temporal coherence across event intervals, and (ii) a geometry-driven shape anomaly detection mechanism to suppress shadows, specular highlights, and noise. The method integrates event-stream parsing, temporal profile modeling, and classical optimization—requiring no training data. Evaluated on real-world event data from 3D-printed objects, our approach reduces normal estimation error by over 40% compared to the deep learning baseline EventPS-FCN, while demonstrating significantly improved robustness against illumination variations and surface reflectance non-idealities.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Tracking, profiling, and countermeasures against them
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Recently, the energy-efficient photometric stereo method using an event camera has been proposed to recover surface normals from events triggered by changes in logarithmic Lambertian reflections under a moving directional light source. However, EventPS treats each event interval independently, making it sensitive to noise, shadows, and non-Lambertian reflections. This paper proposes Photometric Stereo based on Event Interval Profile (PS-EIP), a robust method that recovers pixelwise surface normals from a time-series profile of event intervals. By exploiting the continuity of the profile and introducing an outlier detection method based on profile shape, our approach enhances robustness against outliers from shadows and specular reflections. Experiments using real event data from 3D-printed objects demonstrate that PS-EIP significantly improves robustness to outliers compared to EventPS's deep-learning variant, EventPS-FCN, without relying on deep learning.

Problem

• Robust surface normal recovery from event camera data
• Overcoming sensitivity to noise and non-Lambertian reflections
• Enhancing outlier resistance in photometric stereo methods

Innovation

• Uses time-series profile of event intervals
• Introduces outlier detection via profile shape
• Enhances robustness without deep learning