PyTorchKR Papers

최신 논문

총 1,524편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:1809.00437

Unsupervised Image Super-Resolution using Cycle-in-Cycle Generative Adversarial Networks

We consider the single image super-resolution problem in a more general case that the low-/high-resolution pairs and the down-sampling process are unavailable. Different from traditional super-resolution formulation, the low-resolution input is further degraded by noises and blurring. This complicated setting makes supervised learning and accurate kernel estimation impossible. To solve this problem, we resort to unsupervised learning without paired data, inspired by the recent successful image-to-image translation applications. With generative adversarial networks (GAN) as the basic component, we propose a Cycle-in-Cycle network structure to tackle the problem within three steps. First, the noisy and blurry input is mapped to a noise-free low-resolution space. Then the intermediate image is up-sampled with a pre-trained deep model. Finally, we fine-tune the two modules in an end-to-end manner to get the high-resolution output. Experiments on NTIRE2018 datasets demonstrate that the proposed unsupervised method achieves comparable results as the state-of-the-art supervised models.

arXiv:1808.08509

Efficient Single Image Super Resolution using Enhanced Learned Group Convolutions

Convolutional Neural Networks (CNNs) have demonstrated great results for the single-image super-resolution (SISR) problem. Currently, most CNN algorithms promote deep and computationally expensive models to solve SISR. However, we propose a novel SISR method that uses relatively less number of computations. On training, we get group convolutions that have unused connections removed. We have refined this system specifically for the task at hand by removing unnecessary modules from original CondenseNet. Further, a reconstruction network consisting of deconvolutional layers has been used in order to upscale to high resolution. All these steps significantly reduce the number of computations required at testing time. Along with this, bicubic upsampled input is added to the network output for easier learning. Our model is named SRCondenseNet. We evaluate the method using various benchmark datasets and show that it performs favourably against the state-of-the-art methods in terms of both accuracy and number of computations required.

arXiv:1808.03344

Deep Learning for Single Image Super-Resolution: A Brief Review

Single image super-resolution (SISR) is a notoriously challenging ill-posed problem, which aims to obtain a high-resolution (HR) output from one of its low-resolution (LR) versions. To solve the SISR problem, recently powerful deep learning algorithms have been employed and achieved the state-of-the-art performance. In this survey, we review representative deep learning-based SISR methods, and group them into two categories according to their major contributions to two essential aspects of SISR: the exploration of efficient neural network architectures for SISR, and the development of effective optimization objectives for deep SISR learning. For each category, a baseline is firstly established and several critical limitations of the baseline are summarized. Then representative works on overcoming these limitations are presented based on their original contents as well as our critical understandings and analyses, and relevant comparisons are conducted from a variety of perspectives. Finally we conclude this review with some vital current challenges and future trends in SISR leveraging deep learning algorithms.

arXiv:1808.00043공식 코드 · Papers with Code

The Unreasonable Effectiveness of Texture Transfer for Single Image Super-resolution

While implicit generative models such as GANs have shown impressive results in high quality image reconstruction and manipulation using a combination of various losses, we consider a simpler approach leading to surprisingly strong results. We show that texture loss alone allows the generation of perceptually high quality images. We provide a better understanding of texture constraining mechanism and develop a novel semantically guided texture constraining method for further improvement. Using a recently developed perceptual metric employing "deep features" and termed LPIPS, the method obtains state-of-the-art results. Moreover, we show that a texture representation of those deep features better capture the perceptual quality of an image than the original deep features. Using texture information, off-the-shelf deep classification networks (without training) perform as well as the best performing (tuned and calibrated) LPIPS metrics. The code is publicly available.

arXiv:1807.10806

Gated Fusion Network for Joint Image Deblurring and Super-Resolution

Single-image super-resolution is a fundamental task for vision applications to enhance the image quality with respect to spatial resolution. If the input image contains degraded pixels, the artifacts caused by the degradation could be amplified by super-resolution methods. Image blur is a common degradation source. Images captured by moving or still cameras are inevitably affected by motion blur due to relative movements between sensors and objects. In this work, we focus on the super-resolution task with the presence of motion blur. We propose a deep gated fusion convolution neural network to generate a clear high-resolution frame from a single natural image with severe blur. By decomposing the feature extraction step into two task-independent streams, the dual-branch design can facilitate the training process by avoiding learning the mixed degradation all-in-one and thus enhance the final high-resolution prediction results. Extensive experiments demonstrate that our method generates sharper super-resolved images from low-resolution inputs with high computational efficiency.

arXiv:1807.11458

To learn image super-resolution, use a GAN to learn how to do image degradation first

This paper is on image and face super-resolution. The vast majority of prior work for this problem focus on how to increase the resolution of low-resolution images which are artificially generated by simple bilinear down-sampling (or in a few cases by blurring followed by down-sampling).We show that such methods fail to produce good results when applied to real-world low-resolution, low quality images. To circumvent this problem, we propose a two-stage process which firstly trains a High-to-Low Generative Adversarial Network (GAN) to learn how to degrade and downsample high-resolution images requiring, during training, only unpaired high and low-resolution images. Once this is achieved, the output of this network is used to train a Low-to-High GAN for image super-resolution using this time paired low- and high-resolution images. Our main result is that this network can be now used to efectively increase the quality of real-world low-resolution images. We have applied the proposed pipeline for the problem of face super-resolution where we report large improvement over baselines and prior work although the proposed method is potentially applicable to other object categories.

arXiv:1807.07930

Perceptual Video Super Resolution with Enhanced Temporal Consistency

With the advent of perceptual loss functions, new possibilities in super-resolution have emerged, and we currently have models that successfully generate near-photorealistic high-resolution images from their low-resolution observations. Up to now, however, such approaches have been exclusively limited to single image super-resolution. The application of perceptual loss functions on video processing still entails several challenges, mostly related to the lack of temporal consistency of the generated images, i.e., flickering artifacts. In this work, we present a novel adversarial recurrent network for video upscaling that is able to produce realistic textures in a temporally consistent way. The proposed architecture naturally leverages information from previous frames due to its recurrent architecture, i.e. the input to the generator is composed of the low-resolution image and, additionally, the warped output of the network at the previous step. Together with a video discriminator, we also propose additional loss functions to further reinforce temporal consistency in the generated sequences. The experimental validation of our algorithm shows the effectiveness of our approach which obtains images with high perceptual quality and improved temporal consistency.

arXiv:1807.06779

An Attention-Based Approach for Single Image Super Resolution

The main challenge of single image super resolution (SISR) is the recovery of high frequency details such as tiny textures. However, most of the state-of-the-art methods lack specific modules to identify high frequency areas, causing the output image to be blurred. We propose an attention-based approach to give a discrimination between texture areas and smooth areas. After the positions of high frequency details are located, high frequency compensation is carried out. This approach can incorporate with previously proposed SISR networks. By providing high frequency enhancement, better performance and visual effect are achieved. We also propose our own SISR network composed of DenseRes blocks. The block provides an effective way to combine the low level features and high level features. Extensive benchmark evaluation shows that our proposed method achieves significant improvement over the state-of-the-art works in SISR.

arXiv:1807.02758공식 코드 · Papers with Code

Image Super-Resolution Using Very Deep Residual Channel Attention Networks

Convolutional neural network (CNN) depth is of crucial importance for image super-resolution (SR). However, we observe that deeper networks for image SR are more difficult to train. The low-resolution inputs and features contain abundant low-frequency information, which is treated equally across channels, hence hindering the representational ability of CNNs. To solve these problems, we propose the very deep residual channel attention networks (RCAN). Specifically, we propose a residual in residual (RIR) structure to form very deep network, which consists of several residual groups with long skip connections. Each residual group contains some residual blocks with short skip connections. Meanwhile, RIR allows abundant low-frequency information to be bypassed through multiple skip connections, making the main network focus on learning high-frequency information. Furthermore, we propose a channel attention mechanism to adaptively rescale channel-wise features by considering interdependencies among channels. Extensive experiments show that our RCAN achieves better accuracy and visual improvements against state-of-the-art methods.

arXiv:1806.11475

SynNet: Structure-Preserving Fully Convolutional Networks for Medical Image Synthesis

Cross modal image syntheses is gaining significant interests for its ability to estimate target images of a different modality from a given set of source images,like estimating MR to MR, MR to CT, CT to PET etc, without the need for an actual acquisition.Though they show potential for applications in radiation therapy planning,image super resolution, atlas construction, image segmentation etc.The synthesis results are not as accurate as the actual acquisition.In this paper,we address the problem of multi modal image synthesis by proposing a fully convolutional deep learning architecture called the SynNet.We extend the proposed architecture for various input output configurations. And finally, we propose a structure preserving custom loss function for cross-modal image synthesis.We validate the proposed SynNet and its extended framework on BRATS dataset with comparisons against three state-of-the art methods.And the results of the proposed custom loss function is validated against the traditional loss function used by the state-of-the-art methods for cross modal image synthesis.

arXiv:1806.09074

CT-image Super Resolution Using 3D Convolutional Neural Network

Computed Tomography (CT) imaging technique is widely used in geological exploration, medical diagnosis and other fields. In practice, however, the resolution of CT image is usually limited by scanning devices and great expense. Super resolution (SR) methods based on deep learning have achieved surprising performance in two-dimensional (2D) images. Unfortunately, there are few effective SR algorithms for three-dimensional (3D) images. In this paper, we proposed a novel network named as three-dimensional super resolution convolutional neural network (3DSRCNN) to realize voxel super resolution for CT images. To solve the practical problems in training process such as slow convergence of network training, insufficient memory, etc., we utilized adjustable learning rate, residual-learning, gradient clipping, momentum stochastic gradient descent (SGD) strategies to optimize training procedure. In addition, we have explored the empirical guidelines to set appropriate number of layers of network and how to use residual learning strategy. Additionally, previous learning-based algorithms need to separately train for different scale factors for reconstruction, yet our single model can complete the multi-scale SR. At last, our method has better performance in terms of PSNR, SSIM and efficiency compared with conventional methods.

arXiv:1204.5309

Analysis Operator Learning and Its Application to Image Reconstruction

Exploiting a priori known structural information lies at the core of many image reconstruction methods that can be stated as inverse problems. The synthesis model, which assumes that images can be decomposed into a linear combination of very few atoms of some dictionary, is now a well established tool for the design of image reconstruction algorithms. An interesting alternative is the analysis model, where the signal is multiplied by an analysis operator and the outcome is assumed to be the sparse. This approach has only recently gained increasing interest. The quality of reconstruction methods based on an analysis model severely depends on the right choice of the suitable operator. In this work, we present an algorithm for learning an analysis operator from training images. Our method is based on an $\ell_p$-norm minimization on the set of full rank matrices with normalized columns. We carefully introduce the employed conjugate gradient method on manifolds, and explain the underlying geometry of the constraints. Moreover, we compare our approach to state-of-the-art methods for image denoising, inpainting, and single image super-resolution. Our numerical results show competitive performance of our general approach in all presented applications compared to the specialized state-of-the-art techniques.

arXiv:1304.7153

A Convex Approach for Image Hallucination

In this paper we propose a global convex approach for image hallucination. Altering the idea of classical multi image super resolution (SU) systems to single image SU, we incorporate aligned images to hallucinate the output. Our work is based on the paper of Tappen et al. where they use a non-convex model for image hallucination. In comparison we formulate a convex primal optimization problem and derive a fast converging primal-dual algorithm with a global optimal solution. We use a database with face images to incorporate high-frequency details to the high-resolution output. We show that we can achieve state-of-the-art results by using a convex approach.

arXiv:1309.2057

Single image super resolution in spatial and wavelet domain

Recently single image super resolution is very important research area to generate high resolution image from given low resolution image. Algorithms of single image resolution are mainly based on wavelet domain and spatial domain. Filters support to model the regularity of natural images is exploited in wavelet domain while edges of images get sharp during up sampling in spatial domain. Here single image super resolution algorithm is presented which based on both spatial and wavelet domain and take the advantage of both. Algorithm is iterative and use back projection to minimize reconstruction error. Wavelet based denoising method is also introduced to remove noise.

arXiv:1402.3926

Sparse Coding Approach for Multi-Frame Image Super Resolution

An image super-resolution method from multiple observation of low-resolution images is proposed. The method is based on sub-pixel accuracy block matching for estimating relative displacements of observed images, and sparse signal representation for estimating the corresponding high-resolution image. Relative displacements of small patches of observed low-resolution images are accurately estimated by a computationally efficient block matching method. Since the estimated displacements are also regarded as a warping component of image degradation process, the matching results are directly utilized to generate low-resolution dictionary for sparse image representation. The matching scores of the block matching are used to select a subset of low-resolution patches for reconstructing a high-resolution patch, that is, an adaptive selection of informative low-resolution images is realized. When there is only one low-resolution image, the proposed method works as a single-frame super-resolution method. The proposed method is shown to perform comparable or superior to conventional single- and multi-frame super-resolution methods through experiments using various real-world datasets.

arXiv:1404.2999

A Reverse Hierarchy Model for Predicting Eye Fixations

A number of psychological and physiological evidences suggest that early visual attention works in a coarse-to-fine way, which lays a basis for the reverse hierarchy theory (RHT). This theory states that attention propagates from the top level of the visual hierarchy that processes gist and abstract information of input, to the bottom level that processes local details. Inspired by the theory, we develop a computational model for saliency detection in images. First, the original image is downsampled to different scales to constitute a pyramid. Then, saliency on each layer is obtained by image super-resolution reconstruction from the layer above, which is defined as unpredictability from this coarse-to-fine reconstruction. Finally, saliency on each layer of the pyramid is fused into stochastic fixations through a probabilistic model, where attention initiates from the top layer and propagates downward through the pyramid. Extensive experiments on two standard eye-tracking datasets show that the proposed method can achieve competitive results with state-of-the-art models.

arXiv:1402.5564

Structure Tensor Based Image Interpolation Method

Feature preserving image interpolation is an active area in image processing field. In this paper a new direct edge directed image super-resolution algorithm based on structure tensors is proposed. Using an isotropic Gaussian filter, the structure tensor at each pixel of the input image is computed and the pixels are classified to three distinct classes; uniform region, corners and edges, according to the eigenvalues of the structure tensor. Due to application of the isotropic Gaussian filter, the classification is robust to noise presented in image. Based on the tangent eigenvector of the structure tensor, the edge direction is determined and used for interpolation along the edges. In comparison to some previous edge directed image interpolation methods, the proposed method achieves higher quality in both subjective and objective aspects. Also the proposed method outperforms previous methods in case of noisy and JPEG compressed images. Furthermore, without the need for optimization in the process, the algorithm can achieve higher speed.

arXiv:1503.03630

Single image super-resolution by approximated Heaviside functions

Image super-resolution is a process to enhance image resolution. It is widely used in medical imaging, satellite imaging, target recognition, etc. In this paper, we conduct continuous modeling and assume that the unknown image intensity function is defined on a continuous domain and belongs to a space with a redundant basis. We propose a new iterative model for single image super-resolution based on an observation: an image is consisted of smooth components and non-smooth components, and we use two classes of approximated Heaviside functions (AHFs) to represent them respectively. Due to sparsity of the non-smooth components, a $L_{1}$ model is employed. In addition, we apply the proposed iterative model to image patches to reduce computation and storage. Comparisons with some existing competitive methods show the effectiveness of the proposed method.