Learning Compact Hyperbolic Representations of Latent Space for Old Photo Restoration

Recent restoration methods for handling real old photos have achieved significant improvements using generative networks. However, the restoration quality under the usual generative architectures is greatly affected by the encoded properties of latent space, which reflect pivotal semantic information in the recovery process. Therefore, how to find the suitable latent space and identify its semantic factors is an important issue in this challenging task. To this end, we propose a novel generative network with hyperbolic embeddings to restore old photos that suffer from multiple degradations. Specifically, we transform high-dimensional Euclidean features into a compact latent space via the hyperbolic operations. In order to enhance the hierarchical representative capability, we perform the channel mixing and group convolutions for the intermediate hyperbolic features. By using attention-based aggregation mechanism in a hyperbolic space, we can further obtain the resulting latent vectors, which are more effective in encoding the important semantic factors and improving the restoration quality. Besides, we design a diversity loss to guide each latent vector to disentangle different semantics. Extensive experiments have shown that our method is able to generate visually pleasing photos and outperforms state-of-the-art restoration methods.

I. Introduction

Old photo restoration aims to recover the valuable contents from one real degraded photo. As the important memories of beautiful moments in the life, old photo prints are very possible to deteriorate when long-term stored in poor environment. However, it is very challenging to automatically repair damaged old photos due to severe degradation factors such as scratches, blurriness, film noise and color fading. Previous attempts [1], [2] restore the digitalized photos by detecting the localized defects and inpainting the damaged areas. Yet these methods focus on completing the missing contents and are difficult to handle the photos with complex defects. In the era of deep learning, to reconstruct the corrupted regions of old photos, this typical ill-posed problem is generally solved by learning the right feature mappings between the degraded photos and clean ones.