Dictionary Learning Step Research Articles

Iterative Learning Based Modulating Functions Method for Distributed Solar Source Estimation

Modulating functions method is a non asymptotic estimation method, which provides accurate and robust estimations of states, parameters and inputs for different classes of systems, which include unknown linear ordinary differential systems, fractional systems and linear partial differential equations. In the case of time or space varying unknown, the method requires the decomposition of the unknown into predefined basis functions. However, the estimation performance will depend on the nature of the basis functions which in some cases are not easy to determine. This letter proposes a new iterative learning based modulating functions method, which combines the standard modulating functions with a dictionary learning procedure. The dictionary learning step allows the determination of appropriate set of functions to decompose the unknown, while the modulating function step allows the non-asymptotic and robust estimation of the projection coefficients. The performance of the proposed method is illustrated in a distributed solar collector application, modeled by partial differential equations and where the unknown solar irradiance is estimated.

Superpixel based compression of hyperspectral image with modified dictionary and sparse representation

ABSTRACT Sparse representation provides an efficient way for the compression of hyperspectral images in the literature. In this work, an improved version of the Spectral-Spatial Adaptive Sparse Representation (SSASR), Modified SSASR (MSSASR), is proposed for hyperspectral image compression. In the first step of the proposed method, superpixel maps are generated for adaptive spatio-spectral representation. Then, the best possible dictionary is computed for the representation of the data. Afterwards, sparse coefficients are determined depending on the dictionary by Simultaneous Orthogonal Matching Pursuit (SOMP) method. In the final step, the dictionary and sparse coefficients are encoded by quantization and entropy encoding. This paper has the following novelties: modified dictionary learning step, new ordering scheme and Differential Pulse Code Modulation (DPCM) usage. Owing to modified dictionary learning, the sparse coefficients can be represented more compact than traditional SSASR. By using of new ordering scheme, it is not needed to send the superpixel map as side information. Moreover, DCPM usage lowers the magnitudes of sparse coefficients. Thanks to these modifications, the proposed method achieves an important improvement on compression performance. In the experimental results, the proposed method is compared with PCA+JPEG2000, DWT+JPEG2000, 3D-SPECK, 3D-TARP and SSASR methods on Indian Pines, Washington DC Mall, Jasper Ridge and Moffett Field scenes. The evaluation is carried out not only using distance and similarity metrics, namely, signal-to-noise ratio, mean spectral angle and mean spectral correlation metrics but also computation times. Additionally, reconstruction quality in anomaly regions is also used for comparison. Experimental results show that the proposed method outperforms the other compression methods in terms of quality metrics and anomaly preserving performance.

T2-FDL: A robust sparse representation method using adaptive type-2 fuzzy dictionary learning for medical image classification

In this paper, a robust sparse representation for medical image classification is proposed based on the adaptive type-2 fuzzy learning (T2-FDL) system. In the proposed method, sparse coding and dictionary learning processes are executed iteratively until a near-optimal dictionary is obtained. The sparse coding step aiming at finding a combination of dictionary atoms to represent the input data efficiently, and the dictionary learning step rigorously adjusts a minimum set of dictionary items. The two-step operation helps create an adaptive sparse representation algorithm by involving the type-2 fuzzy sets in the design process of image classification. Since the existing image measurements are not made under the same conditions and with the same accuracy, the performance of medical diagnosis is always affected by noise and uncertainty. By introducing an adaptive type-2 fuzzy learning method, a better approximation in an environment with higher degrees of uncertainty and noise is achieved. The experiments are executed over two open-access brain tumor magnetic resonance image databases, REMBRANDT and TCGA-LGG, from The Cancer Imaging Archive (TCIA). The experimental results of a brain tumor classification task show that the proposed T2-FDL method can adequately minimize the negative effects of uncertainty in the input images. The results demonstrate the outperformance of T2-FDL compared to other important classification methods in the literature, in terms of accuracy, specificity, and sensitivity.

Robust dictionary learning for erratic noise-corrupted seismic data reconstruction

Dictionary learning methods adaptively train their bases from the given data in an iterative manner; hence, they can capture more detailed features and achieve sparser representation than a method that uses a fixed basis. However, there also exists a good chance of erratic noise corrupting the dictionary because of the insufficiency of the L1-norm regularization in distinguishing the signal and erratic noise, causing the inaccuracy of both dictionary learning and its application. A robust dictionary learning method is proposed to reconstruct the missing data even in the presence of strong erratic noise. Specifically, a projected operator, which is constructed with the robust Huber misfit, is used to damp erratic noise to a low-amplitude level. In the dictionary learning step, erratic noise is gradually reduced to an acceptable level with the help of this projected operator, from which the signal prototypes (or atoms) can be safely extracted in each iteration. Then, the learned dictionary is used to iteratively estimate the signal from the noisy and under-sampled data, performing noise attenuation as well as data interpolation at non-recording locations. We test the proposed dictionary learning method using 2D and 3D noisy seismic examples and compare it with other state-of-the-art methods. Numerical results demonstrate its superior effectiveness at recovering missing data and increasing signal-to-noise ratio in the presence of erratic noise.

Discovering and characterizing dynamic functional brain networks in task FMRI

Many existing studies for the mapping of function brain networks impose an implicit assumption that the networks' spatial distributions are constant over time. However, the latest research reports reveal that functional brain networks are dynamical and have time-varying spatial patterns. Furthermore, how these functional networks evolve over time has not been elaborated and explained in sufficient details yet. In this paper, we aim to discover and characterize the dynamics of functional brain networks via a windowed group-wise dictionary learning and sparse coding approach. First, we aggregated the sampled subjects' fMRI signals into one big data matrix, and learned a common dictionary for all individuals via a group-wise dictionary learning step. Second, we obtained the dynamic time-varying functional networks by using the windowed time-varying sparse coding approach. Experimental results demonstrated that our windowed group-wise dictionary learning and sparse coding method can effectively detect the task-evoked networks and also characterize how these networks evolve over time. This work sheds novel insights on the dynamics mechanism of functional brain networks.

Reducing the Computational Complexity of EEG Source Localization With Cortical Patch Decomposition and Optimal Electrode Selection.

Real-time implementation of EEG source localization can be employed in a broad area of applications such as clinical diagnosis of neurologic diseases and brain-computer interface. However, a power-efficient, low-complexity, and real-time implementation of EEG source localization is still challenging due to extensive iterations in the solutions. In this study, two techniques are introduced to reduce the computational burden of the subspace-based MUltiple SIgnal Classification (MUSIC) algorithm. To shrink the exhaustive search inherent in MUSIC, the cortex is parsed into cortical regions. A novel nomination procedure involving a dictionary learning step will pick a number of regions to be searched for the active sources. In addition, a new electrode selection algorithm based on the Cramer-Rao bound of the errors is introduced to pick the best set of an arbitrary number of electrodes out of the total. The performance of the proposed techniques were evaluated using simulated EEG signal under variation of different parameters such as the number of nominated regions, the signal to noise ratio, and the number of electrodes. The proposed techniques can reduce the computational complexity by up to $90\%$. Furthermore, the proposed techniques were tested on EEG data from an auditory oddball experiment. A good concordance was observed in the comparison of the topographies and the localization errors derived from the proposed technique and regular MUSIC. Such reduction can be exploited in the real-time, long-run, and mobile monitoring of cortical activity for clinical diagnosis and research purposes.

Motion Estimation in Echocardiography Using Sparse Representation and Dictionary Learning.

This paper introduces a new method for cardiac motion estimation in 2-D ultrasound images. The motion estimation problem is formulated as an energy minimization, whose data fidelity term is built using the assumption that the images are corrupted by multiplicative Rayleigh noise. In addition to a classical spatial smoothness constraint, the proposed method exploits the sparse properties of the cardiac motion to regularize the solution via an appropriate dictionary learning step. The proposed method is evaluated on one data set with available ground-truth, including four sequences of highly realistic simulations. The approach is also validated on both healthy and pathological sequences of in vivo data. We evaluate the method in terms of motion estimation accuracy and strain errors and compare the performance with state-of-the-art algorithms. The results show that the proposed method gives competitive results for the considered data. Furthermore, the in vivo strain analysis demonstrates that meaningful clinical interpretation can be obtained from the estimated motion vectors.This paper introduces a new method for cardiac motion estimation in 2-D ultrasound images. The motion estimation problem is formulated as an energy minimization, whose data fidelity term is built using the assumption that the images are corrupted by multiplicative Rayleigh noise. In addition to a classical spatial smoothness constraint, the proposed method exploits the sparse properties of the cardiac motion to regularize the solution via an appropriate dictionary learning step. The proposed method is evaluated on one data set with available ground-truth, including four sequences of highly realistic simulations. The approach is also validated on both healthy and pathological sequences of in vivo data. We evaluate the method in terms of motion estimation accuracy and strain errors and compare the performance with state-of-the-art algorithms. The results show that the proposed method gives competitive results for the considered data. Furthermore, the in vivo strain analysis demonstrates that meaningful clinical interpretation can be obtained from the estimated motion vectors.

Graph regularized nonnegative sparse coding using incoherent dictionary for approximate nearest neighbor search

In this paper, we consider the problem of approximate nearest neighbor (ANN) retrieval with the method of sparse coding, which is a promising tool of discovering compact representation of high-dimensional data. A new study, exploiting the indices of the active set of sparse coded data as retrieval code, exhibits an appealing ANN route. Here our work aims to enhance the method via considering its shortages of the local structure of the data. Our primary innovation is two-fold: We introduce the graph Laplacian regularization to preserve the local structure of the original data into reduced space, which is indeed beneficial to ANN. And we impose non-negativity constraints such that the retrieval code can more effectively reflect the neighborhood relation, thereby cutting down on unnecessary hash collision. To this end, we learn an incoherent dictionary with both rules via a novel formulation of sparse coding. The resulting optimization problem can be provided with an available solution by the widely used iterative scheme, where we resort to the feature-sign search algorithm in the sparse coding step and exploit the method that uses a Lagrange dual for dictionary learning step. Experimental results on publicly available image data sets manifest that the rules are positive to promote the classification and ANN accuracies. Compared with several state-of-the-art ANN techniques, our methods can achieve an interesting improvement in retrieval accuracy.

Compressed sensing magnetic resonance imaging based on dictionary updating and block‐matching and three‐dimensional filtering regularisation

Compressed sensing (CS) enables that magnetic resonance (MR) images can be exactly reconstructed from undersampled k-space data by exploiting the sparsity of MR images in some analytical sparsifying transform or some dictionary. Recent methods are exploiting adaptive patch-based dictionaries for image recovery by alternating between dictionary learning step and image reconstruction step. In this study, the authors propose a novel MR image reconstruction algorithm utilising dictionary updating, which consists of three steps: sparse coding, dictionary updating and image reconstruction. In the dictionary updating step, they perform a first-order series expansion for dictionary–coefficient matrix product via recursive method, and propose an efficient method to solve the new dictionary updating problem. To improve the reconstruction quality, the proposed block-matching and three-dimensional (3D) filtering regularisation is incorporated into the authors’ image CS recovery, which can combine the self-similarities within the image, the 3D transform sparsity and the local sparsity into image recovery process. Experimental simulations demonstrate their proposed algorithms can obtain better reconstruction quality than the previous CS algorithms.

Dictionary learning for data recovery in positron emission tomography

Compressed sensing (CS) aims to recover images from fewer measurements than that governed by the Nyquist sampling theorem. Most CS methods use analytical predefined sparsifying domains such as total variation, wavelets, curvelets, and finite transforms to perform this task. In this study, we evaluated the use of dictionary learning (DL) as a sparsifying domain to reconstruct PET images from partially sampled data, and compared the results to the partially and fully sampled image (baseline).A CS model based on learning an adaptive dictionary over image patches was developed to recover missing observations in PET data acquisition. The recovery was done iteratively in two steps: a dictionary learning step and an image reconstruction step. Two experiments were performed to evaluate the proposed CS recovery algorithm: an IEC phantom study and five patient studies. In each case, 11% of the detectors of a GE PET/CT system were removed and the acquired sinogram data were recovered using the proposed DL algorithm. The recovered images (DL) as well as the partially sampled images (with detector gaps) for both experiments were then compared to the baseline. Comparisons were done by calculating RMSE, contrast recovery and SNR in ROIs drawn in the background, and spheres of the phantom as well as patient lesions.For the phantom experiment, the RMSE for the DL recovered images were 5.8% when compared with the baseline images while it was 17.5% for the partially sampled images. In the patients’ studies, RMSE for the DL recovered images were 3.8%, while it was 11.3% for the partially sampled images. Our proposed CS with DL is a good approach to recover partially sampled PET data. This approach has implications toward reducing scanner cost while maintaining accurate PET image quantification.

Overlapping Sound Event Detection Using NMF with K-SVD Based Dictionary Learning

Non-Negative Matrix Factorization (NMF) is a method for updating dictionary and gain in alternating manner. Due to ease of implementation and intuitive interpretation, NMF is widely used to detect and separate overlapping sound events. However, NMF that utilizes non-negativity constraints generates parts-based representation and this distinct property leads to a dictionary containing fragmented acoustic events. As a result, the presence of shared basis results in performance degradation in both separation and detection tasks of overlapping sound events. In this paper, we propose a new method that utilizes K-Singular Value Decomposition (K-SVD) based dictionary to address and mitigate the part-based representation issue during the dictionary learning step. Subsequently, we calculate the gain using NMF in sound event detection step. We evaluate and confirm that overlapping sound event detection performance of the proposed method is better than the conventional method that utilizes NMF based dictionary.

Automatic stellar spectral classification via sparse representations and dictionary learning

Stellar classification is an important topic in astronomical tasks such as the study of stellar populations. However, stellar classification of a region of the sky is a time-consuming process due to the large amount of objects present in an image. Therefore, automatic techniques to speed up the process are required. In this work, we study the application of a sparse representation and a dictionary learning for automatic spectral stellar classification. Our dataset consist of 529 calibrated stellar spectra of classes B to K, belonging to the Pulkovo Spectrophotometric catalog, in the 3400−5500A range. These stellar spectra are used for both training and testing of the proposed methodology. The sparse technique is applied by using the greedy algorithm OMP (Orthogonal Matching Pursuit) for finding an approximated solution, and the K-SVD (K-Singular Value Decomposition) for the dictionary learning step. Thus, sparse classification is based on the recognition of the common characteristics of a particular stellar type through the construction of a trained basis. In this work, we propose a classification criterion that evaluates the results of the sparse representation techniques and determines the final classification of the spectra. This methodology demonstrates its ability to achieve levels of classification comparable with automatic methodologies previously reported such as the Maximum Correlation Coefficient (MCC) and Artificial Neural Networks (ANN).

Pixel to Patch Sampling Structure and Local Neighboring Intensity Relationship Patterns for Texture Classification

In this letter, we explore local image descriptors for texture classification. We mainly propose two novel contributions: an effective sampling structure based on Pixel To Patch (PTP) to mimic the retinal sampling pattern; and a novel Local Neighboring Intensity Relationship Pattern (LNIRP) descriptor to extract texture feature by exploring neighboring gray-scale properties. The LNIRP descriptor is extended by using the PTP sampling structure which aims to capture not only micro-patterns but also macro-patterns, while reducing feature dimensionality and improving computational efficiency. The proposed descriptor has advantages of computational simplicity, no texton dictionary learning step and training-free. Moreover, the LNIRP descriptor is complementary to the Local Binary Pattern (LBP) descriptor. Extensive experiments were conducted on Outex database to evaluate the proposed descriptor and sampling structure. The proposed descriptor can achieve superior classification performance compared to most of the state-of-the-art methods, including what we believe to be the best results reported for Outex, while offering a smallest feature dimension.