Transparent Shape from a Single View Polarization Image

TL;DR

This paper introduces a learning-based approach for estimating transparent surfaces from a single polarization image, overcoming challenges of transmission interference by leveraging a physics-inspired confidence mechanism.

Contribution

It proposes a novel physics-based prior and a multi-branch neural network architecture for improved transparent shape estimation from polarization images.

Findings

Achieves superior accuracy compared to existing methods.

Effectively handles transmission interference in polarization images.

Provides a new dataset for transparent shape from polarization.

Abstract

This paper presents a learning-based method for transparent surface estimation from a single view polarization image. Existing shape from polarization(SfP) methods have the difficulty in estimating transparent shape since the inherent transmission interference heavily reduces the reliability of physics-based prior. To address this challenge, we propose the concept of physics-based prior, which is inspired by the characteristic that the transmission component in the polarization image has more noise than reflection. The confidence is used to determine the contribution of the interfered physics-based prior. Then, we build a network(TransSfP) with multi-branch architecture to avoid the destruction of relationships between different hierarchical inputs. To train and test our method, we construct a dataset for transparent shape from polarization with paired polarization images and ground-truth normal maps. Extensive experiments and comparisons demonstrate the superior accuracy of our method.