使用上下文损失保持自然图像统计信息

Mar, 2018

Learning to Maintain Natural Image Statistics

Roey Mechrez, Itamar Talmi, Firas Shama, Lihi Zelnik-Manor

TL;DR本文提出了一种用于训练 feed-forward CNN 的互补方法，目标是维护自然内部特征分布，并可用于GAN。该方法大幅减少了训练所需的图片数量，并取得了单张图像超分辨率和高分辨率表面法线估计的最先进结果。

Abstract

Maintaining natural image statistics is a crucial factor in restoration and generation of realistic looking images. When training cnns, photoreal