Deep Sampling Networks

TL;DR

Deep Sampling Networks (DSN) introduce a novel approach for image down-sampling and up-sampling that avoids traditional interpolation, enhancing super-resolution and image compression performance through learned, end-to-end trainable subnetworks.

Contribution

The paper proposes a deep sampling network with separate subnetworks for down-sampling and up-sampling, eliminating the need for traditional interpolation and achieving state-of-the-art results.

Findings

DSN's down-sampling produces photo-realistic low-resolution images.

DSN achieves superior super-resolution performance.

DSN improves image compression when integrated with codecs.

Abstract

Deep convolutional neural networks achieve excellent image up-sampling performance. However, CNN-based methods tend to restore high-resolution results highly depending on traditional interpolations (e.g. bicubic). In this paper, we present a deep sampling network (DSN) for down-sampling and up-sampling without any cheap interpolation. First, the down-sampling subnetwork is trained without supervision, thereby preserving more information and producing better visual effects in the low-resolution image. Second, the up-sampling subnetwork learns a sub-pixel residual with dense connections to accelerate convergence and improve performance. DSN's down-sampling subnetwork can be used to generate photo-realistic low-resolution images and replace traditional down-sampling method in image processing. With the powerful down-sampling process, the co-training DSN set a new state-of-the-art performance for image super-resolution. Moreover, DSN is compatible with existing image codecs to improve image compression.