The applicability of compressive sensing (CS) to radar imaging has been recently proven and its capability to construct reliable radar images from a limited set of measurements demonstrated. In this study, a common fr...
详细信息
The applicability of compressive sensing (CS) to radar imaging has been recently proven and its capability to construct reliable radar images from a limited set of measurements demonstrated. In this study, a common framework for inverse synthetic aperture radar (ISAR) imaging via CS is provided and a CS-based ISAR imaging method is proposed. The proposed method is tested for application such as imagereconstructionfrom compressed data, resolution enhancement and imagereconstructionfrom gapped data. The effectiveness of the proposed method is demonstrated on real datasets and the performance evaluated by means of image contrast.
Statistical imagereconstruction (SR) algorithms have the potential to significantly reduce x-ray CT image artefacts because they use a more accurate model than conventional filtered backprojection and can incorporate...
详细信息
Statistical imagereconstruction (SR) algorithms have the potential to significantly reduce x-ray CT image artefacts because they use a more accurate model than conventional filtered backprojection and can incorporate effects such as noise, incompletedata and nonlinear detector response. Most SR algorithms assume that the CT detectors are photon-counting devices and generate Poisson-distributed signals. However, actual CT detectors integrate energy from the x-ray beam and exhibit compound Poisson-distributed signal statistics. This study presents the first assessment of the impact on image quality of the resultant mismatch between the detector and signal statistics models assumed by the sinogram data model and the reconstruction algorithm. A 2D CT projection simulator was created to generate synthetic polyenergetic transmission data assuming (i) photon-counting with simple Poisson- distributed signals and (ii) energy-weighted detection with compound Poisson- distributed signals. An alternating minimization (AM) algorithm was used to reconstruct images from the data models (i) and (ii) for a typical abdominal scan protocol with incident particle fluence levels ranging from 105 to 1.6 x 10(6) photons/detector. The images reconstructed fromdata models (i) and (ii) were compared by visual inspection and image-quality figures of merit. The reconstructed image quality degraded significantly when the means were mismatched from the assumed model. However, if the signal means are appropriately modified, images fromdata models (i) and (ii) did not differ significantly even when SNR is very low. While data-mean mismatches characteristic of the difference between particle-fluence and energy-fluence transmission can cause significant streaking and cupping artefacts, the mismatch between the actual and assumed CT detector signal statistics did not significantly degrade image quality once systematic datameans mismatches were corrected.
In industrial X-ray computerized tomography (CT), the objects to be inspected are usually very attenuating to X-rays, and their shape may not permit complete scannings at all view angles; incomplete-data imaging situa...
详细信息
In industrial X-ray computerized tomography (CT), the objects to be inspected are usually very attenuating to X-rays, and their shape may not permit complete scannings at all view angles; incomplete-data imaging situations usually result. In earlier reports, it was concluded that imagereconstructionfromincompletedata can be achieved effectively through an iterative transform algorithm, which utilizes the a priori information on the object to compensate for the missing data. The image is repeatedly transformed to the object space, where it is corrected by the a priori information on the object, and back to the projection space where it is corrected by the measured projections. The results of validating the iterative transform algorithm on experimental datafrom a cross section of a high-pressure turbine blade made of Ni-based superalloy are reported. from the data set, two kinds of incompletedata situations are simulated: incomplete projection and limited-angle scanning. The results indicate that substantial improvements, both visually and in wall thickness measurements, were brought about in all cases through the use of the iterative transform algorithm.
An estimation approach to three-dimensional reconstructionfrom parallel ray projections, with incomplete and very noisy data, is described. Using a stochastic dynamic model for an object of interest in a probed domai...
详细信息
An estimation approach to three-dimensional reconstructionfrom parallel ray projections, with incomplete and very noisy data, is described. Using a stochastic dynamic model for an object of interest in a probed domain of known background density, the reconstruction problem is reformulated as a n onlinear state estimation problem. An approximate minimum mean square error globally optimal algorithm for the solution of this problem is presented. The algorithm, which is recursive in a hybrid frequency-space domain, operates directly on the Fourier transformed projection data, eliminating altogether the attempt to invert the projection integral equation. The simulation example considered in this paper demonstrates that good object estimates may be obtained with as few as five views in a limited sector of 90° and at a signal-to-noise ratio as low as 0 dB.
X-ray differential phase-contrast computed tomography (DPC-CT) is a powerful physical and biochemical analysis tool. In practical applications, there are often challenges for DPC-CT due to insufficient data caused by ...
详细信息
X-ray differential phase-contrast computed tomography (DPC-CT) is a powerful physical and biochemical analysis tool. In practical applications, there are often challenges for DPC-CT due to insufficient data caused by few-view, bad or missing detector channels, or limited scanning angular range. They occur quite frequently because of experimental constraints from imaging hardware, scanning geometry, and the exposure dose delivered to living specimens. In this work, we analyze the influence of incompletedata on DPC-CT imagereconstruction. Then, a reconstruction method is developed and investigated for incompletedata DPC-CT. It is based on an algebraic iteration reconstruction technique, which minimizes the image total variation and permits accurate tomographic imaging with less data. This work comprises a numerical study of the method and its experimental verification using a dataset measured at the W2 beamline of the storage ring DORIS iiI equipped with a Talbot-Lau interferometer. The numerical and experimental results demonstrate that the presented method can handle incompletedata. It will be of interest for a wide range of DPC-CT applications in medicine, biology, and nondestructive testing.
Nowadays, pervasive vision through synthetic aperture radar (SAR) imaging sensors is a crucial part of air-borne and space-borne platforms to reach the concept of internet of multimedia things over satellites and visu...
详细信息
Nowadays, pervasive vision through synthetic aperture radar (SAR) imaging sensors is a crucial part of air-borne and space-borne platforms to reach the concept of internet of multimedia things over satellites and visual flying ad-hoc networks. SAR sensors always need high-performance datareconstruction techniques to provide quality of experience for end-users of the distributed surveillance nodes and ensuring reliable decision-making in the autonomous vehicles. Using prior information in SAR imagereconstruction, improves the quality of reconstructed image. Many sparse regularization methods use pre information terms in which the image is sparsely presented based on a predefined dictionary. However, if the desired features of the real image have not a sparse representation based on the predefined dictionary, imagereconstruction with enhanced interested features will be failed. Dictionary learning to better adapt with the underlying scene can lead to better imagereconstruction. On the other hand, using the idea of dictionary learning in the imagereconstruction problem, and considering that the data used to train the dictionary may be incomplete and noisy, will create limitations for this method. In this paper, a new idea based on the use of an overcomplete dictionary consisting of a learned dictionary and a predefined dictionary (pre-learned dictionary) for SAR imagereconstruction problem is presented. Developing the necessary mathematical relations and providing the framework for SAR imagereconstruction based on the use of the overcomplete pre-learned dictionary, to simultaneously enhance the features considered in advance and the features associated with the image, fromincomplete and noisy SAR raw data is presented in this article. We also present an iterative algorithm for solving the corresponding optimization problem. Simulation results based on the real data of TerraSAR-X and the Lynx airborne system show the effectiveness of the proposed method.
In this article, we propose an unsupervised deep learning method for positron emission tomography (PET) reconstructionfromincompletedata. This method utilizes the so-called deep image prior (DIP) as an untrained de...
详细信息
In this article, we propose an unsupervised deep learning method for positron emission tomography (PET) reconstructionfromincompletedata. This method utilizes the so-called deep image prior (DIP) as an untrained deep convolutional neural network (CNN) to generate object reconstructions. The main idea is to reparameterize the imagereconstruction problem as a neural network optimization problem. We show that the proposed method effectively addresses the incompletedatareconstruction problem, which otherwise degrades the image resolution and quality of standard reconstruction algorithms. Meanwhile, the proposed method does not require any pretraining procedures, i.e., it is not biased toward any particular dataset. Hence, it has the potential to be used in clinical situations, where training data would be infeasible or prohibitively expensive. The performance of the proposed approach is evaluated with noisy synthetic data based on the Shepp-Logan and brainweb phantoms, and clinical naive rat data. In addition, robustness studies of the approach with respect to regularization parameters are also carried out. We showcase that the proposed method considerably outperforms the state-of-the-art methods, leading to flexible reconstructionfromincomplete PET data.
Practical methods for imagereconstructionfrom hollow and from truncated projections are presented. The incomplete projections are preprocessed so that reconstructions can be obtained from them using the modified bac...
详细信息
Practical methods for imagereconstructionfrom hollow and from truncated projections are presented. The incomplete projections are preprocessed so that reconstructions can be obtained from them using the modified back-projection method. There are two types of preprocessing procedure. In the first, we use a smooth continuation of each hollow or truncated projection so that the augmented data has the form of a complete projection. This preprocessing approach is more appropriate for truncated than for hollow projections. The second preprocessing procedure uses a smooth continuation of all the projections simultaneously by operating on their angular Fourier coefficients. Conditions are imposed which ensure that the augmented projection data is consistent with an object of finite extent. This preprocessing approach is of limited value for truncated projections but it enables accurate reconstructions to be obtained when the projections are hollow. Examples showing reconstructions of a test object from computer-generated, incomplete projections are presented (thereby showing the effectiveness of the preprocessing methods) in a companion paper.
imagereconstructionfromincomplete measurements is one basic task in imaging. While supervised deep learning has emerged as a powerful tool for imagereconstruction in recent years, its applicability is limited by i...
详细信息
imagereconstructionfromincomplete measurements is one basic task in imaging. While supervised deep learning has emerged as a powerful tool for imagereconstruction in recent years, its applicability is limited by its prerequisite on a large number of latent images for model training. To extend the application of deep learning to the imaging tasks where acquisition of latent images is challenging, this article proposes an unsupervised deep learning method that trains a deep model for imagereconstruction with the access limited to measurement data. We develop a Siamese network whose twin sub-networks perform reconstruction cooperatively on a pair of complementary spaces: the null space of the measurement matrix and the range space of its pseudo inverse. The Siamese network is trained by a self-supervised loss with three terms: a data consistency loss over available measurements in the range space, a data consistency loss between intermediate results in the null space, and a mutual consistency loss on the predictions of the twin sub-networks in the full space. The proposed method is applied to four imaging tasks from different applications, and extensive experiments have shown its advantages over existing unsupervised solutions.
暂无评论