Search

In recent years, deep convolutional neural networks (CNNs) have been widely applied to handle low-level vision problems. However, most existing CNN-based approaches can either handle single degeneration each time or treat them jointly through feature entangling, thus likely leading to poor performance when the actual degradation is inconsistent with hypothetical degradation condition. Furthermore, feature coupling will bring a large amount of computation, which may make the methods impractical to real-time mobile scenarios. In order to address these problems, we propose a deep decoupled cooperative learning model which can not only develop the corresponding recover network to deal with each degradation, but also flexibly handle multiple degradations at the same time. Thus, our approach can achieve disentangling and synthesizing single image super-resolution and motion deblurring, which has high practicability. We evaluate the proposed approach on various benchmark datasets, covering both natural images and synthetic images. The results demonstrate its superiority, compared to the state-of-the-art, where image SR and motion deblurring can be accomplished effectively concurrently. The source code of the work is available at https://github.com/hengliusky/Cooperative-Learning-Deblur-SR .

Satellite imagery is a kind of typical remote sensing data, which holds preponderance in large area imaging and strong macro integrity. However, for most commercial space usages, such as virtual display of urban traffic flow, virtual interaction of environmental resources, one drawback of satellite imagery is its low spatial resolution, failing to provide the clear image details. Moreover, in recent years, synthesizing the color for grayscale satellite imagery or recovering the original color of camouflage sensitive regions becomes an urgent requirement for large spatial objects virtual reality interaction. In this work, unlike existing works which solve these two problems separately, we focus on achieving image super-resolution (SR) and image colorization synchronously. Based on multi-task learning, we provide a novel deep neural network model to fulfill single satellite imagery SR and colorization simultaneously. By feeding back the color feature representations into the SR network and jointly optimizing such two tasks, our deep model successfully achieves the mutual cooperation between imagery reconstruction and image colorization. To avoid color bias, we not only adopt the non-satellite imagery to enrich the color diversity of satellite image, but also recalculate the prior color distribution and the valid color range based on the mixed data. We evaluate the proposed model on satellite images from different data sets, such as RSSCN7 and AID. Both the evaluations and comparisons reveal that the proposed multi-task deep learning approach is superior to the state-of-the-art methods, where image SR and colorization can be accomplished simultaneously and efficiently.

This paper presents an end-to-end multi-scale deep encoder (convolution) and decoder (deconvolution) network for single image super-resolution (SISR) guided by phase congruency (PC) edge map. Our system starts by a single scale symmetrical encoder–decoder structure for SISR, which is extended to a multi-scale model by integrating wavelet multi-resolution analysis into our network. The new multi-scale deep learning system allows the low resolution (LR) input and its PC edge map to be combined so as to precisely predict the multi-scale super-resolved edge details with the guidance of the high-resolution (HR) PC edge map. In this way, the proposed deep model takes both the reconstruction of image pixels’ intensities and the recovery of multi-scale edge details into consideration under the same framework. We evaluate the proposed model on benchmark datasets of different data scenarios, such as Set14 and BSD100 - natural images, Middlebury and New Tsukuba - depth images. The evaluations based on both PSNR and visual perception reveal that the proposed model is superior to the state-of-the-art methods.