Panchromatic image based dictionary learning for hyperspectral imagery denoising

As a dual-modal imaging technology that has emerged in recent years, cone-beam X-ray luminescence computed tomography (CB-XLCT) has exhibited promise as a tool for the early three-dimensional detection of tumors in small animals. However, due to the challenges imposed by the low absorption and high scattering of light in tissues, the CB-XLCT reconstruction problem is a severely ill-conditioned inverse problem, rendering it difficult to obtain satisfactory reconstruction results. In this study, a strategy that utilizes dictionary learning and group structure (DLGS) is proposed to achieve satisfactory CB-XLCT reconstruction performance. The group structure is employed to account for the clustering of nanophosphors in specific regions within the organism, which can enhance the interrelation of elements in the same group. Furthermore, the dictionary learning strategy is implemented to effectively capture sparse features. The performance of the proposed method was evaluated through numerical simulations and in vivo experiments. The experimental results demonstrate that the proposed method achieves superior reconstruction performance in terms of location accuracy, target shape, robustness, dual-source resolution, and in vivo practicability.

Hyperspectral imagery (HSI) denoising is an important preprocessing step for real-world applications. Recently, sparse representation and low-rank representation based methods are proven effective in HSI denoising. However, most of these approaches only consider the low-rankness in the spectral domain and the sparsity in coding matrix. They have ignored the property that the coding matrix of each atom is also low-rank, i.e., low-rankness also exists in the spatial domain. In this paper, a reweighed sparse low-rank nonnegative tensor factorization (RSLRNTF) method is proposed to restore an HSI. It takes an HSI as a third-order tensor and factorizes it into the combination of a few component tensors where each one is the outer product of a low-rank matrix (coding matrix) and a vector (atom). Additionally, a reweighed L1 norm is added to coding matrices to enforce their sparsity. The low-rankness in both the spatial domain and the spectral domain as well as sparsity in the spatial domain improve the denoising performance. Furthermore, the nonnegativity in both coding matrices and dictionary leads to parts-based representation of HSI, which facilitates preserving local fine structure information. Experimental results on synthetic data and real-world data demonstrate the superiority of proposed method.

Efficient denoising of hyperspectral imagery (HSI) relies on an representational model that is capable of capturing the spatial and spectral correlation in HSI. Recently, an intrinsic representation (IR) approach based on the linear spectral mixture model (LSMM) was proposed for unsupervised feature extraction. The IR model constitutes a sound representational model due to its ability to account for the physical data generation process of HSI, the spatial correlation effect, and the noise variance heterogeneity effect. In this paper, we explore the potential of IR for the denoising of HSI. A noisy pixel in HSI is expressed as a nonnegative linear combination of several endmembers, plus some Gaussian noise with heterogeneous noise variances. In order to perform denoising, the IR approach is used to adaptively estimate both the endmembers and the nonnegative coefficients (i.e., the abundances), which are finally used to reconstruct the clean image. The experiments on both simulated and real hyperspectral images demonstrate that the IR approach not only can resist the influence of noise, but also can preserve the image details.

Non-overlapping classification of hyperspectral imagery with superpixel segmentation

This paper proposes an efficient and high-fidelity image fusion method based on adaptive smoothing filtering for panchromatic (PAN) and multispectral (MS) image fusion. The scale ratio reflects the ratio of spatial resolution between the panchromatic image and the multispectral image. When facing a multiscale fusion task, traditional methods are unable to simultaneously handle the problems of spectral resolution loss resulting from high scale ratios and the issue of reduced spatial resolution due to low scale ratios. To adapt to the fusion of panchromatic and multispectral satellite images of different scales, this paper improves the problem of the insufficient filtering of high-frequency information of remote sensing images of different scales by the classic filter-based intensity modulation (SFIM) model. It uses Gaussian convolution kernels instead of traditional mean convolution kernels and builds a Gaussian pyramid to adaptively construct convolution kernels of different scales to filter out high-frequency information of high-resolution images. It can adaptively process panchromatic multispectral images of different scales, iteratively filter the spatial information in panchromatic images, and ensure that the scale transformation is consistent with the definition of multispectral images. Using 15 common fusion methods, this paper compares the experimental results of ZY-3 with scale ratio 2.7 and SV-1 with scale ratio 4 data. The results show that the method proposed in this paper retains good spatial information for image fusion at different scales and has good spectral preservation.

Hyperspectral imagery (HSI) denoising is a popular research topic in remote sensing. In this paper, we propose a novel method for HSI denoising by performing Minimum Noise Fraction (MNF) to the original HSI data cube, thresholding the noisy output bands with the Video Non-Local Bayes (VNLB) algorithm, and then conducting the inverse MNF transform to obtain the denoised data cube. Our experiments demonstrate that the proposed method usually achieves the best denoising results among several existing denoising methods for two HSI data cubes. In addition, it is much faster for HSI denoising than the VNLB algorithm which was originally developed for video denoising.

Low-rank double dictionary learning from corrupted data for robust image classification

This paper analyzes the role of dictionary selection in Sparse Representation-based Classification (SRC). While SRC introduces interesting results in the field of classification, its performance is highly limited by the number of training samples to form the classification matrix. Different studies addressed this issue by using a more compact representation of the training data in order to achieve higher classification speed and accuracy. Representative selection methods which are analyzed in this paper include Metaface dictionary learning, Fisher Discriminative Dictionary Learning (FDDL), Sparse Modeling Representative Selection (SMRS), and random selection of the training samples. The first two methods build their own dictionaries via an optimization process while the other two methods select the representatives directly from the original training samples. These methods, along with the original method which uses all training samples to form the classification matrix, were examined on two face datasets and one digit dataset. The role of feature extraction was also studied using two dimensionality reduction methods, down-sampling and random projection. The results show that the FDDL method leads to the best classification accuracy followed by the SMRS method as the second best. On the other hand, the SMRS method requires a much smaller learning time which makes it more appropriate for dynamic situations where the dictionary is regularly updated with new samples. The accuracy of the Metaface dictionary learning method was specifically less than the other two methods. As expected, using all the training samples as the dictionary resulted in the best recognition rates in all the datasets but the classification times for this approach were far larger than the required time using any of the three dictionary learning methods.

To improve the magnetic resonance imaging (MRI) data acquisition speed while maintaining the reconstruction quality, a novel method is proposed for multislice MRI reconstruction from undersampled k-space data based on compressed-sensing theory using dictionary learning. There are two aspects to improve the reconstruction quality. One is that spatial correlation among slices is used by extending the atoms in dictionary learning from patches to blocks. The other is that the dictionary-learning scheme is used at two resolution levels; i.e., a low-resolution dictionary is used for sparse coding and a high-resolution dictionary is used for image updating. Numerical experiments are carried out on in vivo 3D MR images of brains and abdomens with a variety of undersampling schemes and ratios. The proposed method (dual-DLMRI) achieves better reconstruction quality than conventional reconstruction methods, with the peak signal-to-noise ratio being 7 dB higher. The advantages of the dual dictionaries are obvious compared with the single dictionary. Parameter variations ranging from 50% to 200% only bias the image quality within 15% in terms of the peak signal-to-noise ratio. Dual-DLMRI effectively uses the a priori information in the dual-dictionary scheme and provides dramatically improved reconstruction quality.

In this letter, a novel nonlinear supervised dictionary learning (DL) scheme called multiscale incremental DL, whose objective function contains reconstruction error terms and a classification error term, is proposed for synthetic aperture radar (SAR) object recognition. In the reconstruction error terms, considering the local and global features of SAR images, Gaussian functions with different blurring parameters are exploited to extract SAR images' multiscale features, and all features can be reconstructed according to the weights assigned to these features at different scales. In the classification error term, a linear combination of classification vectors close to the labels of samples restricts sparse codes from different classes to be almost independent. Furthermore, an incremental method is utilized to address the memory consumption problem, and the optimal solution is obtained. Experiments on the moving and stationary target automatic recognition database demonstrate that the proposed algorithm outperforms several representative DL, support vector machine, and k-nearest neighbor methods in the case of a small training sample set size and exhibits strong antinoise performance.

In this paper, we conduct a study for the effects of principal component analysis (PCA) on hyperspectral imagery denoising. Our previous works combined PCA with wavelet shrinkage, block matching and 3D filtering (BM3D), and block matching and 4D filtering (BM4D), respectively, and very good denoising results have been obtained for hyperspectral imagery with very little noise. To demonstrate if these methods are the best for other noise scenarios, we combine PCA with video BM3D (VBM3D) and video BM4D (VBM4D), and non-local means (NL-Means) as well. Experimental results show that the PCA+VBM3D is the best denoising method for moderate and high noise levels, and PCA+BM4D is preferred for very low noise levels. KeywordsHyperspectral imagery denoisingBM3DBM4DVBM3DVBM4DPrincipal component analysis (PCA)

Dictionary learning (DL) technique has received a great interest recently, due to its significant role in feature extraction. Although many DL‐based methods have been presented, some of them still suffer from the lack of discriminative features, especially for the local manifold features. To mitigate this problem, the authors propose a novel DL method named low‐rank representation based on twin tensor kernel (LRR‐TTK) DL for face recognition in this study. Specifically, the training samples are projected to a high‐dimensional space with TTK. Then, they extract the local manifold features and spatial features (representation coefficients) hidden in the facial images by TT locality preserving projection. In addition, powered by LRR reconstruction and DL theory, much more discriminative features are obtained, which can improve the recognition rate greatly. Comprehensive experimental results at AR, extended Yale‐B and FERET face databases demonstrate the superiority of their proposed method.

Dynamic cerebral perfusion x-ray computed tomography (PCT) imaging has been advocated to quantitatively and qualitatively assess hemodynamic parameters in the diagnosis of acute stroke or chronic cerebrovascular diseases. However, the associated radiation dose is a significant concern to patients due to its dynamic scan protocol. To address this issue, in this paper we propose an image restoration method by utilizing coupled dictionary learning (CDL) scheme to yield clinically acceptable PCT images with low-dose data acquisition. Specifically, in the present CDL scheme, the 2D background information from the average of the baseline time frames of low-dose unenhanced CT images and the 3D enhancement information from normal-dose sequential cerebral PCT images are exploited to train the dictionary atoms respectively. After getting the two trained dictionaries, we couple them to represent the desired PCT images as spatio-temporal prior in objective function construction. Finally, the low-dose dynamic cerebral PCT images are restored by using a general DL image processing. To get a robust solution, the objective function is solved by using a modified dictionary learning based image restoration algorithm. The experimental results on clinical data show that the present method can yield more accurate kinetic enhanced details and diagnostic hemodynamic parameter maps than the state-of-the-art methods.

This paper presents a robust blind separation method of sparse sources named robust sparse blind source separation (r-SBSS). Allowing for the morphological diversity of the source, we first assume that one source can be divide into two parts, namely global component and local component. Furthermore, a novel optimization model is built upon the low-rank matrix recovery and the sparse modeling of both the sources and the outliers with different regularized constraints. We also provide the corresponding optimization algorithm, which allows combining both low-rank double dictionary learning and low-rank representation strategy to realize source separation and mixing matrix estimation respectively. In the proposed method, the dictionary learning strategy guarantees the adaptive representation of sparsity, and the low-rank representation strategy is first introduced to implement the effective estimation of the mixing matrix and outliers. Extensive experimental results on AR dataset and real EMI signals demonstrate the superiority of r-SBSS, compared to the state-of-the-art methods.

Due to the data-hungry nature of deep neural networks and the high dimensionality of hyperspectral imagery (HSI) data, the scarcity of training samples remains a significant challenge for HSI classification based on deep learning. Recently, the pairing or recombining of samples has proven to be an efficient approach to increase the amount of input training data, and accordingly, the architecture of a network trained by these samples has to be redesigned, since the input is no longer a single sample. Following this strategy, in this article, we propose a spatial–spectral relation network (SS-RN) for HSI classification with limited training samples. The SS-RN takes advantage of the relation network architecture to precisely capture the profound similarity between samples. Additionally, to make SS-RN more suitable for HSI classification, an entire three-dimensional (3-D) neighborhood instead of the isolated pixel is considered to explore both spectral and spatial information thoroughly. Multiple support samples are also selected for each class to make sure that the extracted features are stable enough to avoid the intraclass similarity and interclass dissimilarity problem of HSI. The network is composed on the basis of 3-D convolutional neural network blocks with the aim of extracting spatial–spectral features. Extensive experiments on three widely used HSI datasets demonstrate that the proposed method can achieve better classification accuracy than conventional deep learning methods with limited labeled training samples.