Preservation Of Image Features Research Articles

A dense and U-shaped transformer with dual-domain multi-loss function for sparse-view CT reconstruction.

CT image reconstruction from sparse-view projections is an important imaging configuration for low-dose CT, as it can reduce radiation dose. However, the CT images reconstructed from sparse-view projections by traditional analytic algorithms suffer from severe sparse artifacts. Therefore, it is of great value to develop advanced methods to suppress these artifacts. In this work, we aim to use a deep learning (DL)-based method to suppress sparse artifacts. Inspired by the good performance of DenseNet and Transformer architecture in computer vision tasks, we propose a Dense U-shaped Transformer (D-U-Transformer) to suppress sparse artifacts. This architecture exploits the advantages of densely connected convolutions in capturing local context and Transformer in modelling long-range dependencies, and applies channel attention to fusion features. Moreover, we design a dual-domain multi-loss function with learned weights for the optimization of the model to further improve image quality. Experimental results of our proposed D-U-Transformer yield performance improvements on the well-known Mayo Clinic LDCT dataset over several representative DL-based models in terms of artifact suppression and image feature preservation. Extensive internal ablation experiments demonstrate the effectiveness of the components in the proposed model for sparse-view computed tomography (SVCT) reconstruction. The proposed method can effectively suppress sparse artifacts and achieve high-precision SVCT reconstruction, thus promoting clinical CT scanning towards low-dose radiation and high-quality imaging. The findings of this work can be applied to denoising and artifact removal tasks in CT and other medical images.

A variational model to remove multiplicative noise based on SAR image feature preservation

A variational model to remove multiplicative noise based on SAR image feature preservation

A Quantized Convolutional Neural Network Implemented With Memristor for Image Denoising and Recognition.

The interference of noise will cause the degradation of image quality, which can have a negative impact on the subsequent image processing and visual effect. Although the existing image denoising algorithms are relatively perfect, their computational efficiency is restricted by the performance of the computer, and the computational process consumes a lot of energy. In this paper, we propose a method for image denoising and recognition based on multi-conductance states of memristor devices. By regulating the evolution of Pt/ZnO/Pt memristor wires, 26 continuous conductance states were obtained. The image feature preservation and noise reduction are realized via the mapping between the conductance state and the image pixel. Furthermore, weight quantization of convolutional neural network is realized based on multi-conductance states. The simulation results show the feasibility of CNN for image denoising and recognition based on multi-conductance states. This method has a certain guiding significance for the construction of high-performance image noise reduction hardware system.

Variational Bayesian Learning of SMoGs: Modelling and Their Application to Synthetic Aperture Radar

We show how modern Bayesian Machine Learning tools can be effectively used in order to develop efficient methods for filtering Earth Observation signals. Bayesian statistical methods can be thought of as a generalization of the classical least-squares adjustment methods where both the unknown signals and the parameters are endowed with probability distributions, the priors. Statistical inference under this scheme is the derivation of posterior distributions, that is, distributions of the unknowns after the model has seen the data. Least squares can then be thought of as a special case that uses Gaussian likelihoods, or error statistics. In principle, for most non-trivial models, this framework requires performing integration in high-dimensional spaces. Variational methods are effective tools for approximate inference in Statistical Machine Learning and Computational Statistics. In this paper, after introducing the general variational Bayesian learning method, we apply it to the modelling and implementation of sparse mixtures of Gaussians (SMoG) models, intended to be used as adaptive priors for the efficient representation of sparse signals in applications such as wavelet-type analysis. Wavelet decomposition methods have been very successful in denoising real-world, non-stationary signals that may also contain discontinuities. For this purpose we construct a constrained hierarchical Bayesian model capturing the salient characteristics of such sets of decomposition coefficients. We express our model as a Dirichlet mixture model. We then show how variational ideas can be used to derive efficient methods for bypassing the need for integration: the task of integration becomes one of optimization. We apply our SMoG implementation to the problem of denoising of Synthetic Aperture Radar images, inherently affected by speckle noise, and show that it achieves improved performance compared to established methods, both in terms of speckle reduction and image feature preservation.

Face Recognition for Varying Illumination and Different Optical Zoom Using a Combination of Binary and Geometric Features

Preservation of image features caused by binary conversion is a difficult task under variation of illumination conditions. Several binary conversion-based methods have used an adaptive thresholding technique to improve their performance under illumination variation conditions because of its robustness. However, the performances of existing methods were still limited under high differences illumination conditions especially for uncontrolled lighting sources. In addition, various length of face-to-camera distance gives significant problem affect for the performance of face recognition method. It happens when various images are available for the same person with different length face-to-camera distances due to the appearance of varying facial features of the same person. Therefore, this study proposed to combine the strength normalization and feature-based method to build an illumination distribution model to overcome this problem. With the proposed method the illumination model will fit with variation of illumination conditions in a whole image to generate an adaptive threshold for a novel columnar binary conversion method. The proposed method consists of five main stages, starting with eye area detection using the developed Viola–Jones algorithm. Next, the iris is detected using the Circular Hough Transform (CHT) method and will convert it into binary using the proposed Columnar Binary Conversion (CBC) method to preserve the appearance of the facial features under the illumination variation. Then, the proposed Facial Feature Region Normalization (FFRN) method is performed to improve the effects of different optical zooms for the classification step. The classification is conducted based on the similarity measurement between the extracted normalised binary face region and the dataset that must be converted into their equivalent normalised binary images. The proposed method is evaluated on two different smartphone databases, namely as Smartphone Face Video (SFV) and MOBIO. The performance results showed the outperformance of the proposed method.

A dual-stream deep convolutional network for reducing metal streak artifacts in CT images

Machine learning and deep learning are rapidly finding applications in the medical imaging field. In this paper, we address the long-standing problem of metal artifacts in computed tomography (CT) images by training a dual-stream deep convolutional neural network for streak removal. While many metal artifact reduction methods exist, even state-of-the-art algorithms fall short in some clinical applications. Specifically, proton therapy planning requires high image quality with accurate tumor volumes to ensure treatment success. We explore a dual-stream deep network structure with residual learning to correct metal streak artifacts after a first-pass by a state-of-the-art interpolation-based algorithm, NMAR. We provide the network with a mask of the streaks in order to focus attention on those areas. Our experiments compare a mean squared error loss function with a perceptual loss function to emphasize preservation of image features and texture. Both visual and quantitative metrics are used to assess the resulting image quality for metal implant cases. Success may be due to the duality of information processing, with one network stream performing local structure correction, while the other stream provides an attention mechanism to destreak effectively. This study shows that image-domain deep learning can be highly effective for metal artifact reduction (MAR), and highlights the benefits and drawbacks of different loss functions for solving a major CT reconstruction challenge.

Model-based myocardial T1 mapping with sparsity constraints using single-shot inversion-recovery radial FLASH cardiovascular magnetic resonance

BackgroundThis study develops a model-based myocardial T1 mapping technique with sparsity constraints which employs a single-shot inversion-recovery (IR) radial fast low angle shot (FLASH) cardiovascular magnetic resonance (CMR) acquisition. The method should offer high resolution, accuracy, precision and reproducibility.MethodsThe proposed reconstruction estimates myocardial parameter maps directly from undersampled k-space which is continuously measured by IR radial FLASH with a 4 s breathhold and retrospectively sorted based on a cardiac trigger signal. Joint sparsity constraints are imposed on the parameter maps to further improve T1 precision. Validations involved studies of an experimental phantom and 8 healthy adult subjects.ResultsIn comparison to an IR spin-echo reference method, phantom experiments with T1 values ranging from 300 to 1500 ms revealed good accuracy and precision at simulated heart rates between 40 and 100 bpm. In vivo T1 maps achieved better precision and qualitatively better preservation of image features for the proposed method than a real-time CMR approach followed by pixelwise fitting. Apart from good inter-observer reproducibility (0.6% of the mean), in vivo results confirmed good intra-subject reproducibility (1.05% of the mean for intra-scan and 1.17, 1.51% of the means for the two inter-scans, respectively) of the proposed method.ConclusionModel-based reconstructions with sparsity constraints allow for single-shot myocardial T1 maps with high spatial resolution, accuracy, precision and reproducibility within a 4 s breathhold. Clinical trials are warranted.

A simple and fast adaptive nonlocal multispectral filtering algorithm for efficient noise reduction in magnetic resonance imaging

PurposeWe recently introduced a multispectral (MS) nonlocal (NL) filter based on maximum likelihood estimation (MLE) of voxel intensities, termed MS-NLML. While MS-NLML provides excellent noise reduction and improved image feature preservation as compared to other NL or MS filters, it requires considerable processing time, limiting its application in routine analyses. In this work, we introduced a fast, simple, and robust filter, termed nonlocal estimation of multispectral magnitudes (NESMA), for noise reduction in multispectral (MS) magnetic resonance imaging (MRI). MethodsThrough extensive simulation and in-vivo analyses, we compared the performance of NESMA and MS-NLML in terms of noise reduction and processing efficiency. Further, we introduce two simple adaptive methods that permit spatial variation of similar voxels, R, used in the filtering. The first method is semi-adaptive and permits variation of R across the image by using a relative Euclidean distance (RED) similarity threshold. The second method is fully adaptive and filters the raw data with several RED similarity thresholds to spatially determine the optimal threshold value using an unbiased criterion. ResultsNESMA shows very similar filtering performance as compared to MS-NLML, however, with much simple implementation and very fast processing time. Further, for both filters, the adaptive methods were shown to further reduce noise in comparison with the conventional non-adaptive method in which R is set to a constant value throughout the image. ConclusionsNESMA is fast, robust, and straightforward to implement filter. These features render it suitable for routine clinical use and analysis of large MRI datasets.

Reduction of variable-truncation artifacts from beam occlusion during in situ x-ray tomography

Many in situ x-ray tomography studies require experimental rigs which may partially occlude the beam and cause parts of the projection data to be missing. In a study of fluid flow in porous chalk using a percolation cell with four metal bars drastic streak artifacts arise in the filtered backprojection (FBP) reconstruction at certain orientations. Projections with non-trivial variable truncation caused by the metal bars are the source of these variable-truncation artifacts. To understand the artifacts a mathematical model of variable-truncation data as a function of metal bar radius and distance to sample is derived and verified numerically and with experimental data. The model accurately describes the arising variable-truncation artifacts across simulated variations of the experimental setup. Three variable-truncation artifact-reduction methods are proposed, all aimed at addressing sinogram discontinuities that are shown to be the source of the streaks. The ‘reduction to limited angle’ (RLA) method simply keeps only non-truncated projections; the ‘detector-directed smoothing’ (DDS) method smooths the discontinuities; while the ‘reflexive boundary condition’ (RBC) method enforces a zero derivative at the discontinuities. Experimental results using both simulated and real data show that the proposed methods effectively reduce variable-truncation artifacts. The RBC method is found to provide the best artifact reduction and preservation of image features using both visual and quantitative assessment. The analysis and artifact-reduction methods are designed in context of FBP reconstruction motivated by computational efficiency practical for large, real synchrotron data. While a specific variable-truncation case is considered, the proposed methods can be applied to general data cut-offs arising in different in situ x-ray tomography experiments.

Joint reconstruction of multi-channel, spectral CT data via constrained total nuclear variation minimization

We explore the use of the recently proposed ‘total nuclear variation’ (TVN) as a regularizer for reconstructing multi-channel, spectral CT images. This convex penalty is a natural extension of the total variation (TV) to vector-valued images and has the advantage of encouraging common edge locations and a shared gradient direction among image channels. We show how it can be incorporated into a general, data-constrained reconstruction framework and derive update equations based on the first-order, primal-dual algorithm of Chambolle and Pock. Early simulation studies based on the numerical XCAT phantom indicate that the inter-channel coupling introduced by the TVN leads to better preservation of image features at high levels of regularization, compared to independent, channel-by-channel TV reconstructions.

Quantitative evaluation of a single-distance phase-retrieval method applied on in-line phase-contrast images of a mouse lung.

Propagation-based X-ray phase-contrast computed tomography (PBI) has already proven its potential in a great variety of soft-tissue-related applications including lung imaging. However, the strong edge enhancement, caused by the phase effects, often hampers image segmentation and therefore the quantitative analysis of data sets. Here, the benefits of applying single-distance phase retrieval prior to the three-dimensional reconstruction (PhR) are discussed and quantified compared with three-dimensional reconstructions of conventional PBI data sets in terms of contrast-to-noise ratio (CNR) and preservation of image features. The PhR data sets show more than a tenfold higher CNR and only minor blurring of the edges when compared with PBI in a predominately absorption-based set-up. Accordingly, phase retrieval increases the sensitivity and provides more functionality in computed tomography imaging.

Wavelet-based minimal-energy approach to image restoration

The popular mathematical models for digital image restoration, formulated as a minimization problem of certain total energy functionals, give rise to the variational/PDE-based approach to process the input image in the spatial (or physical) domain. In general, the total energy in these models consists of two additive terms, namely: the internal energy for dictating the image quality in terms of image smoothness and image feature preservation, and the external energy for ensuring the output image not to deviate too far from the input image. In formulating the internal energy, a specific energy density function is chosen and applied to the magnitude of the gradient operation for extracting image features to be preserved. In this paper, we replace the gradient operation by some wavelet transform that has better performance in feature extraction and eliminates the need of iterative steps. Thus, the problem of image processing is performed in the wavelet domain instead. By taking advantage of the multi-scale structure of wavelets and the corresponding multi-level singularity detection capability, the proposed approach should facilitate further development of fast and effective algorithms for the variational approach, perhaps even with significant reduction in computational complexity as compared with the traditional approach to digital image restoration.

General Ribbon-Based Thinning Algorithms for Stylus-Generated Images

Thinning algorithms for stylus-generated images are presented which are based on the general ribbon model of stylus-generated images. These algorithms have several advantages over existing thinning algorithms, including the existence of a formal specification of the desired output of a thinning algorithm; the preservation of image features which have been shown to be the most perceptually significant for the human perception of stylus-generated images; and the ability to deal easily with images which contain both stroke and blob objects.

Scratch tracking for 2 in video tape applications

Physical scratches on 2in video tapes cause severe degradation in viewable picture quality. A new algorithm for locating these defects is presented, based on the way in which the data are mapped onto the tape and how the scratches affect corresponding image pixels. Also, the mapping onto tape results in optimal image feature preservation, i.e. only the features affected by the scratch are processed.

A new axial smoothing method based on elastic mapping

New PET scanners have higher axial and in-plane spatial resolutions but at the expense of reduced per plane sensitivity, which prevents the higher to be fully realized. Normally, Gaussian weighted interplane axial smoothing is used to reduce noise. In this study, the authors developed a new algorithm to first elastically map from adjacent planes to an expected plane location, then by simple averaging with the measured image at the corresponding axial plane location to reduce the image noise level. Compared with those images obtained by conventional axial-directional smoothing method, the images of the new method which are based on matching image features of adjacent planes have improved signal-to-noise ratio. To quantify the observation, both simulated cardiac images and real cardiac PET images were studied. The in-plane loss was measured by the effective global Gaussian resolution and the noise reduction was evaluated by the cross-correlation coefficient. Various Hanning reconstruction filters with cutoff frequency=0.5, 0.7, 1.0/spl times/Nyquist frequency and Ramp 1.0 filter were tested on simulated images. Results showed that the new method was robust to various noise levels and indicated larger noise reduction or better image feature preservation than by the conventional method.