Revisiting Disparity from Dual-Pixel Images: Physics-Informed Lightweight Depth Estimation

TL;DR

This paper introduces a lightweight, physics-informed disparity estimation method for dual-pixel images that outperforms existing approaches while using fewer parameters and without requiring specialized DP training data.

Contribution

The authors develop a novel completion-based network that explicitly models disparity constraints and leverages physical properties, enabling effective training on RGB-D datasets and robust disparity estimation.

Findings

Achieved state-of-the-art disparity estimation accuracy.

Reduced system size to one-fifth of conventional methods.

Did not require dual-pixel datasets for training.

Abstract

In this study, we propose a high-performance disparity (depth) estimation method using dual-pixel (DP) images with few parameters. Conventional end-to-end deep-learning methods have many parameters but do not fully exploit disparity constraints, which limits their performance. Therefore, we propose a lightweight disparity estimation method based on a completion-based network that explicitly constrains disparity and learns the physical and systemic disparity properties of DP. By modeling the DP-specific disparity error parametrically and using it for sampling during training, the network acquires the unique properties of DP and enhances robustness. This learning also allows us to use a common RGB-D dataset for training without a DP dataset, which is labor-intensive to acquire. Furthermore, we propose a non-learning-based refinement framework that efficiently handles inherent disparity expansion errors by appropriately refining the confidence map of the network output. As a result, the proposed method achieved state-of-the-art results while reducing the overall system size to 1/5 of that of the conventional method, even without using the DP dataset for training, thereby demonstrating its effectiveness. The code and dataset are available on our project site.