A simple convolutional neural network is proposed in this letter and is trained end-to-end to restore clear images from hazy inputs. The proposed network is generic and agnostic in the sense that it is not designed specifically for image dehazing and, in particular, it has no knowledge of the atmosphere scattering model. Remarkably, this network achieves record-breaking dehazing performance on several standard data sets that are synthesized using the atmosphere scattering model. This surprising finding suggests that there might be a need to rethink the predominant plug-in approach to image dehazing.

Hyperspectral target detection is critical in both military and civilian applications. However, it is a challenging task due to the complexity of background and the limited samples of target in hyperspectral images (HSIs). In this paper, we propose a novel background learning model, called background learning based on target suppression constraint (BLTSC) to characterize high-dimensional spectral vectors. Considering insufficient target samples, the model is trained only on the background spectral samples to accurately learn the background distribution. Then the discrepancy between the reconstructed and original HSIs are examined to spot the targets. To obtain a background training dataset, coarse detection is carried out. However, it is quite difficult to retrieve pure background data. Thus a target suppression constraint is imposed to reduce the impact of suspected target samples on background reconstruction. Experiments on six real HSIs demonstrate that the proposed framework significantly outperforms the current state-of-the-art detection methods and yields higher detection accuracy and lower false alarm rate.

With success of Deep Belief Networks (DBNs) in computer vision, DBN has attracted great attention in hyperspectral classification. Many deep learning based algorithms have been focused on deep feature extraction for classification improvement. Multi-features, such as texture feature, are widely utilized in classification process to enhance classification accuracy greatly. In this paper, a novel hyperspectral classification framework based on an optimal DBN and a novel texture feature enhancement (TFE) is proposed. Through band grouping, sample band selection and guided filtering, the texture features of hyperspectral data are improved. After TFE, the optimal DBN is employed on the hyperspectral reconstructed data for feature extraction and classification. Experimental results demonstrate that the proposed classification framework outperforms some state-of-the-art classification algorithms, and it can achieve outstanding hyperspectral classification performance. Furthermore, our proposed TFE method can play a significant role in improving classification accuracy.

Hyperspectral imagery with very high spectral resolution provides a new insight for subtle nuances identification of similar substances. However, hyperspectral target detection faces significant challenges of intraclass dissimilarity and interclass similarity due to the unavoidable interference caused by atmosphere, illumination, and sensor noise. In order to effectively alleviate these spectral inconsistencies, this paper proposes a novel target detection method without strict assumptions on data distribution based on an unconstrained linear mixture model and deep learning. Our proposed detector firstly reduces interference via a specifically designed deep-learning-based hierarchical denoising autoencoder, and then carries out accurate detection with a two-step subspace projection, aiming at background suppression and target enhancement. Additionally, to generate representative background and reliable target samples required in the detection procedure, an efficient spatial-spectral unified endmember extraction method has been developed. Performance comparison with several state-of-the-art detection methods and further analysis on four real-world hyperspectral images demonstrate the effectiveness and efficiency of our proposed target detector.

Restoring degraded underwater images is a challenging ill-posed problem. The existing prior-based approaches have limited performance in many situations due to the reliance on handcrafted features. In this paper, we propose an effective convolutional neural network (CNN) for underwater image restoration. The proposed network consists of two paralleled branches: a transmission estimation network (T-network) and a global ambient light estimation network (A-network); in particular, the T-network employs cross-layer connection and multi-scale estimation to prevent halo artifacts and to preserve edge features. The estimates produced by these two branches are leveraged to restore the clear image according to the underwater optical imaging model. Moreover, we develop a new underwater image synthesizing method for building the training datasets, which can simulate images captured in various underwater environments. Experimental results based on synthetic and real images demonstrate that our restored underwater images exhibit more natural color correction and better visibility improvement against several state-of-the-art methods.

This article presents a deep residual network-based fusion framework for hyperspectral and LiDAR data. In this framework, three new fusion methods are proposed, which are the residual network-based deep feature fusion (RNDFF), the residual network-based probability reconstruction fusion (RNPRF) and the residual network-based probability multiplication fusion (RNPMF). The three methods use extinction profile (EP), local binary pattern (LBP), and deep residual network. Specifically, EP and LBP features are extracted from two sources and stacked as spatial features. For RNDFF, the deep features of each source are extracted by a deep residual network, and then the deep features are stacked to create the fusion features which are classified by softmax classifier. For RNPRF, the deep features of each source are input to the softmax classifier to obtain the probability matrices, and then the probability matrices are fused by weighted addition to producing the final label assignment. For RNPMF, the probability matrices are fused by array multiplication. Experimental results demonstrate that the classification performance of the proposed methods significantly outperform existing methods in hyperspectral and LiDAR data fusion.