Super-resolution MRI images using Compressive Sensing

In this paper we propose a technique for obtaining a high resolution (HR) image from a single low resolution (LR) image -using joint learning dictionary - on the basis of image statistic research. It suggests that with an appropriate choice of an over-complete dictionary, image patches can be well represented as a sparse linear combination. Medical imaging for clinical analysis and medical intervention is being used for creating visual representations of the interior of a body, as well as visual representation of the function of some organs or tissues (physiology). A number of medical imaging techniques are in use like MRI, CT scan, X-rays and Optical Coherence Tomography (OCT). OCT is one of the new technologies in medical imaging and one of its uses is in ophthalmology where it is being used for analysis of the choroidal thickness in the eyes in healthy and disease states such as age-related macular degeneration, central serous chorioretinopathy, diabetic retinopathy and inherited retinal dystrophies. We have proposed a technique for enhancing the OCT images which can be used for clearly identifying and analyzing the particular diseases. Our method uses dictionary learning technique for generating a high resolution image from a single input LR image. We train two joint dictionaries, one with OCT images and the second with multiple different natural images, and compare the results with previous SR technique. Proposed method for both dictionaries produces HR images which are comparatively superior in quality with the other proposed method of SR. Proposed technique is very effective for noisy OCT images and produces up-sampled and enhanced OCT images.

We explore in this paper an efficient algorithmic solution to single image super-resolution (SR). We propose the gCLSR, namely graph-Constrained Least Squares Regression, to super-resolve a high-resolution (HR) image from a single low-resolution (LR) observation. The basic idea of gCLSR is to learn a projection matrix mapping the LR image patch to the HR image patch space while preserving the intrinsic geometric structure of the original HR image patch manifold. Even if gCLSR resembles other manifold learning-based SR methods in preserving the local geometric structure of HR and LR image patch manifolds, the innovation of gCLSR lies in that it preserves the intrinsic geometric structure of the original HR image patch manifold rather than the LR image patch manifold, which may be contaminated by image degeneration (e.g., blurring, down-sampling and noise). Upon acquiring the projection matrix, the target HR image can be simply super-resolved from a single LR image without the need of HR-LR training pairs, which favors resource-limited applications. Experiments on images from the public database show that gCLSR method can achieve competitive quality as state-of-the-art methods, while gCLSR is much more efficient in computation than some state-of-the-art methods.

Single image super-resolution based on compressive sensing and improved TV minimization sparse recovery

Deep Feature Blend Attention: A New Frontier in Super Resolution Image Generation

Learning based single image super-resolution (SISR) for real-world images has been an active research topic yet a challenging task, due to the lack of paired low-resolution (LR) and high-resolution (HR) training images. Most of the existing unsupervised real-world SISR methods adopt a two-stage training strategy by synthesizing realistic LR images from their HR counterparts first, then training the super-resolution (SR) models in a supervised manner. However, the training of image degradation and SR models in this strategy are separate, ignoring the inherent mutual dependency between downscaling and its inverse upscaling process. Additionally, the ill-posed nature of image degradation is not fully considered. In this paper, we propose an image downscaling and SR model dubbed as SDFlow, which simultaneously learns a bidirectional many-to-many mapping between real-world LR and HR images unsupervisedly. The main idea of SDFlow is to decouple image content and degradation information in the latent space, where content information distribution of LR and HR images is matched in a common latent space. Degradation information of the LR images and the high-frequency information of the HR images are fitted to an easy-to-sample conditional distribution. Experimental results on real-world image SR datasets indicate that SDFlow can generate diverse realistic LR and SR images both quantitatively and qualitatively.

In this paper, we demonstrate simple algorithms that project low resolution (LR) images differing in subpixel shifts on a high resolution (HR) also called super resolution (SR) grid. The algorithms are very effective in accuracy as well as time efficiency. A number of spatial interpolation techniques using nearest neighbor, inverse-distance weighted averages, Radial Basis Functions (RBF) etc. are used in projection. For best accuracy of reconstructing SR image by a factor of two requires four LR images differing in four independent subpixel shifts. The algorithm has two steps: i) registration of low resolution images and (ii) shifting the low resolution images to align with reference image and projecting them on high resolution grid based on the shifts of each low resolution image using different interpolation techniques. Experiments are conducted by simulating low resolution images by subpixel shifts and subsampling of original high resolution image and the reconstructing the high resolution images from the simulated low resolution images. The results of accuracy of reconstruction are compared by using mean squared error measure between original high resolution image and reconstructed image. The algorithm was tested on remote sensing images and found to outperform previously proposed techniques such as Iterative Back Projection algorithm (IBP), Maximum Likelihood (ML) algorithms. The algorithms are robust and are not overly sensitive to the registration inaccuracies.

The recently introduced Compressed Sensing (CS) theory explains how sparse or compressible signals can be reconstructed from far fewer samples than what was previously believed possible. The CS theory has attracted significant attention for applications such as Magnetic Resonance Imaging (MRI) where long acquisition times have been problematic. This is especially true for dynamic MRI applications where high spatio-temporal resolution is needed. For example, in cardiac cine MRI, it is desirable to acquire the whole cardiac volume within a single breath-hold in order to avoid artifacts due to respiratory motion. Conventional MRI techniques do not allow reconstruction of high resolution image sequences from such limited amount of data. Vaswani et al. recently proposed an extension of the CS framework to problems with partially known support (i.e. sparsity pattern). In their work, the problem of recursive reconstruction of time sequences of sparse signals was considered. Under the assumption that the support of the signal changes slowly over time, they proposed using the support of the previous frame as the "known" part of the support for the current frame. While this approach works well for image sequences with little or no motion, motion causes significant change in support between adjacent frames. In this paper, we illustrate how motion estimation and compensation techniques can be used to reconstruct more accurate estimates of support for image sequences with substantial motion (such as cardiac MRI). Experimental results using phantoms as well as real MRI data sets illustrate the improved performance of the proposed technique.

This paper investigated the problem of the watermark algorithm for digital color image. An digital image watermark embedding and blind extraction algorithm is proposed based on edge detection and Compressive Sensing (CS) theory. Firstly, the carrier image is divided into non-overlapping image blocks. Then, the target image blocks suitable for concealing the watermark information are chosen by combining the edge detection algorithm. Finally, watermarks are embedded into target image blocks by utilizing the Singular Value Decomposition (SVD) and CS theory. The experimental result shows that the proposed digital images watermark algorithm has excellent invisibility and robustness performance in some scenarios.

Recently, the Magnetic Resonance Imaging (MRI) images have limited and unsatisfactory resolutions due to various constraints such as physical, technological and economic considerations. Super-resolution techniques can obtain high-resolution MRI images. The traditional methods obtained the resolution enhancement of brain MRI by interpolations, affecting the accuracy of the following diagnose process. The requirement for brain image quality is fast increasing. In this paper, we propose an image super-resolution (SR) method based on overcomplete dictionaries and inherent similarity of an image to recover the high-resolution (HR) image from a single low-resolution (LR) image. We explore the nonlocal similarity of the image to tentatively search for similar blocks in the whole image and present a joint reconstruction method based on compressive sensing (CS) and similarity constraints. The sparsity and self-similarity of the image blocks are taken as the constraints. The proposed method is summarized in the following steps. First, a dictionary classification method based on the measurement domain is presented. The image blocks are classified into smooth, texture and edge parts by analyzing their features in the measurement domain. Then, the corresponding dictionaries are trained using the classified image blocks. Equally important, in the reconstruction part, we use the CS reconstruction method to recover the HR brain MRI image, considering both nonlocal similarity and the sparsity of an image as the constraints. This method performs better both visually and quantitatively than some existing methods.

Time delay estimation (TDE) is a hot research topic in wireless location technology. Compressed sensing (CS) theory has been widely applied to image reconstruction and direction of arrival estimation since it was proposed in 2004. The sparse model can be constructed in time domain for estimating the time delay by using the CS theory. The measurement matrix plays a crucial role in the processing of signal reconstruction which is the core problem of CS theory. Therefore the research in the measurement matrix has becomes a hotspot in recent years. The existing measurement matrix is mainly divided into two categories, i.e., random measurement matrix and deterministic measurement matrix. The performance of random measurement matrix has bottlenecks. Firstly, because of the redundant measurement matrix data, the generation and storage of the random number put forward a high requirement for hardware. Secondly the random matrix can only satisfy the restricted isometry property in a statistical sense. The research of the deterministic measurement matrix is of great value under this background. The parity check matrix of low density parity check (LDPC) code has good performance in CS theory. However, the method of randomly selecting non-zero element position has a certain probability to generate a measurement matrix with a short loop structure during generating LDPC code measurement matrix. The robustness of the reconstruction performance decreases with the increase of iteration times. A novel quasi-cyclic CS algorithm based on progressive edge-growth is constructed to estimate the time delay. The purpose of this article is to deal with the need to store a large number of data in existing measurement matrix during time delay, by using the CS theory. The algorithm presented here can achieve TDE in a high precision. First, the theoretical bridge between CS and the maximum likelihood decoding is established. And the design criterion of measurement matrix based on the LDPC code is derived. The sparse measurement matrix with quasi-cyclic structure is constructed by introducing the idea of progressive edge-growth. Finally, the orthogonal matching pursuit algorithm is used to estimate the time delay. Furthermore, the computational complexity of the algorithm and the data storage of the measurement matrix are analyzed theoretically. Simulations show that the correct reconstruction probability of the proposed approach is higher than those of the Gauss random matrix and random LDPC matrix under the same dimension. Compared with the random LDPC matrix, the proposed method can improve performance at the expense of less complexity under the condition of the same data storage.

Spectrum acquisition of imaging scenes with super spectral resolution can be realized by multichannel spectral camera with broadband absorption filters under the condition that the multichannel spectral camera sensitivity is optimized. Algorithms to broadband absorption filters selection to optimize the camera sensitivity proposed in the past have no strict theoretical guarantees on reconstruction accuracy. Consequently, the insight had not been uncovered until the Compressive Sensing (CS) theory has ripped in the last recent years. Combined the proofed datasets of published literature and sensing matrix design theory of CS algorithm to optimal the sensitivity of multispectral camera by filter selection is proposed and verified. The more variation of filter vectors can be selected, the more accuracy of the spectral reconstruction results can be acquired with super resolution.

Many deep learning (DL) frameworks have demonstrated state-of-the-art performance in the super-resolution (SR) task of magnetic resonance imaging, but most performances have been achieved with simulated low-resolution (LR) images rather than LR images from real acquisition. Due to the limited generalizability of the SR network, enhancement is not guaranteed for real LR images because of the unreality of the training LR images. In this study, we proposed a DL-based SR framework with an emphasis on data construction to achieve better performance on real LR MR images. The framework comprised two steps: (a) downsampling training using a generative adversarial network (GAN) to construct more realistic and perfectly matched LR/high-resolution (HR) pairs. The downsampling GAN input was real LR and HR images. The generator translated the HR images to LR images and the discriminator distinguished the patch-level difference between the synthetic and real LR images. (b) SR training was performed using an enhance4d deep super-resolution network (EDSR). In the controlled experiments, three EDSRs were trained using our proposed method, Gaussian blur, and k-space zero-filling. As for the data, liver MR images were obtained from 24 patients using breath-hold serial LR and HR scans (only HR images were used in the conventional methods). The k-space zero-filling group delivered almost zero enhancement on the real LR images and the Gaussian group produced a considerable number of artifacts. The proposed method exhibited significantly better resolution enhancement and fewer artifacts compared with the other two networks. Our method outperformed the Gaussian method by an improvement of 0.111 ± 0.016 in the structural similarity index and 2.76 ± 0.98 dB in the peak signal-to-noise ratio. The blind/reference-less image spatial quality evaluator metric of the conventional Gaussian method and proposed method were 46.6 ± 4.2 and 34.1 ± 2.4, respectively.

We propose a novel framework to reconstruct super-resolution human shape from a single low-resolution input image. The approach overcomes limitations of existing approaches that reconstruct 3D human shape from a single image, which require high-resolution images together with auxiliary data such as surface normal or a parametric model to reconstruct high-detail shape. The proposed framework represents the reconstructed shape with a high-detail implicit function. Analogous to the objective of 2D image super-resolution, the approach learns the mapping from a low-resolution shape to its high-resolution counterpart and it is applied to reconstruct 3D shape detail from low-resolution images. The approach is trained end-to-end employing a novel loss function which estimates the information lost between a low and high-resolution representation of the same 3D surface shape. Evaluation for single image reconstruction of clothed people demonstrates that our method achieves high-detail surface reconstruction from low-resolution images without auxiliary data. Extensive experiments show that the proposed approach can estimate super-resolution human geometries with a significantly higher level of detail than that obtained with previous approaches when applied to low-resolution images. https://marcopesavento.github.io/SuRS/.

We propose a new model called iterative collaborative representation (ICR) for image super-resolution (SR). Most of popular SR approaches extract low-resolution (LR) features from the given LR image directly to recover its corresponding high-resolution (HR) features. However, they neglect to utilize the reconstructed HR image for further image SR enhancement. Based on this observation, we extract features from the reconstructed HR image to progressively upscale LR image in an iterative way. In the learning phase, we use the reconstructed and the original HR images as inputs to train the mapping models. These mapping models are then used to upscale the original LR images. In the reconstruction phase, mapping models and LR features extracted from the LR and reconstructed image are then used to conduct image SR in each iteration. Experimental results on standard images demonstrate that our ICR obtains state-of-the-art SR performance quantitatively and visually, surpassing recently published leading SR methods.KeywordsIterative collaborative representationSuper-resolution

In this paper we propose a fast approach for edge preserving super-resolution (SR) based on learning of contourlet coefficients. Given a low resolution test image, we first obtain an initial HR estimate i.e., a close approximation to SR image by learning the contourlet coefficients from a training database consisting of low resolution (LR) and high resolution (HR) images. The final SR image is obtained by using a regularization framework in which both the SR and the LR images are modeled as separate homogeneous Markov Random Fields (MRFs). The LR image formation process is modeled as a decimated and noisy version of the SR image and the final cost function is minimized by using a gradient descent method. Novelty of our approach lies in preserving the edges in the final SR image while using a non edge preserving MRF prior. This is definitely advantageous since it avoids the use of discontinuity preserving prior and hence the computationally taxing optimization methods. The edges in the final SR correspond to those learned from the initial HR estimate. The use of MRF on the low resolution image imposes an additional constraint on the final solution and hence we expect a better solution. In addition, we use the initial HR image for estimating the decimation matrix entries as well as for learning the corresponding MRF parameter. We show the effectiveness of the proposed approach by conducting the experiments on images captured using a real camera.