Deep Portrait Delighting

TL;DR

This paper introduces a deep neural network that effectively removes shading from portraits, enhancing image quality and aiding tasks like face relighting and semantic parsing under challenging lighting conditions.

Contribution

The paper proposes a novel training scheme with three regularization strategies for improved portrait delighting and demonstrates its benefits for related computer vision tasks.

Findings

Enhanced delighting quality over state-of-the-art methods

Improved generalization to various lighting conditions

Better performance in light-sensitive vision tasks

Abstract

We present a deep neural network for removing undesirable shading features from an unconstrained portrait image, recovering the underlying texture. Our training scheme incorporates three regularization strategies: masked loss, to emphasize high-frequency shading features; soft-shadow loss, which improves sensitivity to subtle changes in lighting; and shading-offset estimation, to supervise separation of shading and texture. Our method demonstrates improved delighting quality and generalization when compared with the state-of-the-art. We further demonstrate how our delighting method can enhance the performance of light-sensitive computer vision tasks such as face relighting and semantic parsing, allowing them to handle extreme lighting conditions.