OPAL: Occlusion Pattern Aware Loss for Unsupervised Light Field Disparity Estimation

TL;DR

This paper introduces OPAL, an unsupervised light field disparity estimation method that leverages occlusion pattern encoding to achieve high accuracy, robustness, and better generalization than supervised approaches, with reduced network complexity.

Contribution

The paper proposes OPAL, a novel occlusion pattern aware loss for unsupervised disparity estimation, improving accuracy, robustness, and generalization without ground-truth data.

Findings

Significantly outperforms state-of-the-art unsupervised methods in accuracy.

Demonstrates strong generalization to real-world data.

Reduces network parameters for efficient inference.

Abstract

Light field disparity estimation is an essential task in computer vision with various applications. Although supervised learning-based methods have achieved both higher accuracy and efficiency than traditional optimization-based methods, the dependency on ground-truth disparity for training limits the overall generalization performance not to say for real-world scenarios where the ground-truth disparity is hard to capture. In this paper, we argue that unsupervised methods can achieve comparable accuracy, but, more importantly, much higher generalization capacity and efficiency than supervised methods. Specifically, we present the Occlusion Pattern Aware Loss, named OPAL, which successfully extracts and encodes the general occlusion patterns inherent in the light field for loss calculation. OPAL enables: i) accurate and robust estimation by effectively handling occlusions without using any ground-truth information for training and ii) much efficient performance by significantly reducing the network parameters required for accurate inference. Besides, a transformer-based network and a refinement module are proposed for achieving even more accurate results. Extensive experiments demonstrate our method not only significantly improves the accuracy compared with the SOTA unsupervised methods, but also possesses strong generalization capacity, even for real-world data, compared with supervised methods. Our code will be made publicly available.