Design of a residual block that provides a rich set of features while requiring only small numbers of parameters and operations is crucial for the task of single image super resolution. This is especially important in applications with limited power and storage capacity. In this paper, a new multi-domain residual block is proposed in order to generate richer set of features for the task of image super resolution. The proposed residual block consists of two feature generation modules. The first one is a spatial information processing module and the second one is a spectral information processing module. The feature maps obtained by these two feature generation modules are concatenatively fused to obtain block's output. The new residual block is used to build light-weight super resolution networks. Extensive experiments are performed using several benchmark datasets in order to evaluate the performance of the networks using the new multi-domain residual block. It is shown that the use of both the spatial and spectral features enhances the performance of the light-weight super resolution networks.

Generating features representing the textures and structures of an image is very important characteristic of a super resolution network. Morphological operations are the nonlinear mathematical operations that can process signals focusing on their structures and textures. In this paper, we propose a novel residual block to generate and process morphological features and fuse them with the conventional spatial features, in order to produce a very rich and highly representational set of residual feature maps. The proposed residual block is then used in a deep convolutional neural network for the task of image super resolution. It is shown that the capability of the proposed block in generating and using the morphological features can significantly improve the super resolution performance of a deep network. The super resolution network employing the proposed residual block is shown to outperform the state-of-the-art low-complexity image super resolution networks on various benchmark datasets.

Enhancing the quality of the images acquired under the water environments is crucial in many broadcast technologies. As the richness of the features generated by deep underwater image enhancement networks improves, the visual signals with higher qualities can be yielded. In view of this, in this paper, we propose a new deep network for the task of underwater image enhancement, in which the network feature generation process is guided by the prior information obtained from various underwater medium transmission map and atmospheric light estimation methods. Further, in order to obtain high values for different image quality assessment metrics associated with the images produced by the proposed network, we introduce a multi-stage training process for our network. In the first stage, the proposed network is trained with the conventional supervised learning technique, whereas, in the second stage, the training process of the network is carried out by the adversarial learning technique. Finally, in the third stage, the training of the network obtained by the conventional supervised learning is continued by the guidance of the one trained by the adversarial learning technique. In the development of the adversarial learning-based stage of our network, we propose a novel multi-discriminator generative adversarial network, which is able to produce images with more realistic textures and structures. The proposed multi-discriminator generative adversarial network employs the discrimination process between the real and fake data in various underwater environment color spaces. The results of different experimentations show the effectiveness of the proposed scheme in restoring the high-quality images compared to the other state-of-the-art deep underwater image enhancement networks.

The physical process used in CCD cameras for image formation makes it imperative to simultaneously upsample and deblur the captured images. In this paper, we provide an efficient scheme to solve this problem through a nonlinear end-to-end mapping carried out by a novel deep light-weight residual neural network. The proposed network is designed based on two main modules, namely, image upsampling and image deblurring, aimed for carrying out simultaneously the two tasks involved with the problem. The proposed network employs a residual block with a capacity of generating features in multiple receptive fields and enhancing the network's representational capability. The proposed network is extensively experimented using benchmark datasets for image restoration.

The task of image super resolution is crucial in many applications, such as computer vision and medical imaging. Conventionally, the task of image super resolution was carried out by formulating it as a constrained optimization problem and then solving it using suitable numerical techniques. However, after the emergence of deep neural networks, the focus of the researchers in this area has been almost entirely on designing deep convolutional neural network architectures that indeed have provided remarkable performance for the task of image super resolution. Even though unified methods of combining the two approaches has a greater potential of providing a superior performance for the task of image super resolution, with the exception of very few works, not much attention has been paid to develop such a unified method for this task. In this article, we propose a three-prior formulation of the optimization problem for image super resolution and develop an ultralight-weight convolutional neural network for its solution. The effectiveness of the proposed formulation of the optimization problem and ultralight-weight convolution neural network architecture for its solution is demonstrated through extensive experimentations of the proposed scheme on benchmark datasets and comparisons of the results with that of the other state-of-the-art ultralight-weight image super resolution networks.