Image Reconstruction Error Research Articles

Sensitivity Matrix Update Method for Electrical Resistance Tomography Based on Error-Constrained Cross Fusion Residual Attention Network

Abstract Electrical resistance tomography (ERT) is an imaging technique of conductivity distribution. The image reconstruction algorithm based on the empty field sensitivity matrix is widely used because it provides an approximate solution to the complex ERT inverse problems. However, due to the soft field effect, the sensitivity matrix changes with the media distribution, leading to significant errors in image reconstruction. To address this issue, an error-constrained deep learning scheme for bubble flow, namely cross fusion residual attention network (CFRA-Net) is designed to update the sensitivity matrix during gas-liquid-solid fluidized bed operation. CFRA-Net employs a multi-head self-attention mechanism to enhance bubble features in boundary measurements, a multilevel cross fusion module to fuse empty field strength and boundary measurement information, and a novel serial-channel residual spatial attention block to boost the feature extraction capability of the model. A weighted loss function is designed to give more weight to the gas phase distribution region. Additionally, this paper establishes a dataset for gas-liquid-solid three-phase flow, considering background field conductivity variations. Validation and reliability of the proposed method are assessed through ablation, comparison, and noise experiments. The experimental results demonstrate that CFRA-Net achieves rapid updates of the sensitivity matrix, high imaging accuracy, and robust noise immunity.

Self‐supervised binocular depth estimation algorithm with self‐rectification for autonomous driving

AbstractAiming to address the challenge where existing methods struggle to predict accurate disparities for imperfectly rectified stereo images, and that supervised training requires a considerable amount of ground truth, a self‐supervised binocular depth estimation algorithm with self‐rectification for autonomous driving is proposed. Firstly, a subnetwork dedicated to stereo rectification, aiming to estimate the homography between stereo images is developed. This homography facilitates the transformation of stereo image pairs, aligning their corresponding pixels horizontally. Secondly, a foundational self‐supervised framework primarily centred on minimizing errors in stereo image reconstruction, combined with the generative‐adversarial strategy is introduced. Finally, a vertical offset prediction module (VOPM) is incorporated into the basic framework to further enhance the resistance of the stereo matching network to pixel‐level vertical offset errors. Experimental results on the public KITTI dataset for autonomous driving demonstrate the effectiveness of this approach in improving the disparity prediction performance for imperfectly rectified stereo images. Moreover, the self‐supervised training framework exhibits superiority over state‐of‐the‐art methods.

Towards accurate binocular vision of satellites: A cascaded multi-scale pyramid network for stereo matching on satellite imagery

Stereo matching of high-resolution satellite stereo images is a fundamental and challenging task in photogrammetry, geoscience, and remote sensing. Some deep learning-based methods have shown great potential in solving this task recently. However, these methods have not attracted much attention as their accuracy is still unsatisfactory, especially in regions with intractable factors (e.g., occlusions, repetitive patterns, low texture, non-texture, and disparity discontinuities). To tackle these challenging factors and improve disparity accuracy, we propose an end-to-end stereo matching network named Cascaded Multi-Scale Pyramid Network (CMSP-Net). Firstly, to fully utilize the multi-level spatial context information and mitigate the ambiguous matching caused by repetitive patterns, low texture, non-texture, and occlusions, a cost volume pyramid is constructed with multi-scale image features. Then, each cost volume in the pyramid is aggregated with an hourglass structure to exploit the multi-scale context in a single cost volume. Secondly, considering that high-resolution cost volumes contain more detailed information but have small receptive fields, while low-resolution ones have large receptive fields but lack details, we design an Attention-Guided Cost Fusion (AGCF) strategy to fuse pyramid cost volumes, which effectively integrates the advantages of multi-scale cost volumes while mitigating the backfiring caused by semantic inconsistency across different scales. Thirdly, a disparity refinement module is designed to improve overall disparity accuracy further, which incorporates the reconstruction error and edge clues of input satellite stereo images. Extensive experimental results on multiple satellite datasets demonstrate that the proposed method can effectively reduce matching ambiguities and recover sharp disparity discontinuities, achieving clear superiority over existing state-of-the-art methods.

Unveiling tampering traces: Enhancing image reconstruction errors for visualization

Contemporary tampering detection models in digital image forensics often rely on established techniques such as SRM and ELA for revealing evidence of manipulation. Despite their widespread use, these methods often yield unsatisfactory visualizations. In order to enhance the visibility of tampering artifacts, we propose an innovative image representation called Tamper Reconstruction Error (TRE). TRE measures the error between an input image and its reconstructed counterpart using a pre-trained mixed generator. We observed that utilizing a model proficient in computational visual tasks to extract reconstruction errors did not clearly reveal tampering traces in manually manipulated images. To emphasize the more pronounced discrepancies in the reconstruction of tampered images, the image representation TRE is fed into two dedicated extractors designed to capture manipulation features in both the frequency and spatial domains. Throughout the learning process, these extractors adaptively express the essential forgery traces back into spatial domain. Furthermore, to validate the importance of the extracted errors in tampering localization, we introduced a localization annotator. This annotator integrates reconstruction errors at different stages during the encoding and decoding of latent features. Experimental results demonstrate that the integration of extracted features significantly improves the performance of tampering localization, outperforming other state-of-the-art localization frameworks.

GT-HAD: Gated Transformer for Hyperspectral Anomaly Detection.

Hyperspectral anomaly detection (HAD) aims to distinguish between the background and anomalies in a scene, which has been widely adopted in various applications. Deep neural network (DNN)-based methods have emerged as the predominant solution, wherein the standard paradigm is to discern the background and anomalies based on the error of self-supervised hyperspectral image (HSI) reconstruction. However, current DNN-based methods cannot guarantee correspondence between the background, anomalies, and reconstruction error, which limits the performance of HAD. In this article, we propose a novel gated transformer network for HAD (GT-HAD). Our key observation is that the spatial-spectral similarity in HSI can effectively distinguish between the background and anomalies, which aligns with the fundamental definition of HAD. Consequently, we develop GT-HAD to exploit the spatial-spectral similarity during HSI reconstruction. GT-HAD consists of two distinct branches that model the features of the background and anomalies, respectively, with content similarity as constraints. Furthermore, we introduce an adaptive gating unit to regulate the activation states of these two branches based on a content-matching method (CMM). Extensive experimental results demonstrate the superior performance of GT-HAD. The original code is publicly available at https://github.com/jeline0110/ GT-HAD, along with a comprehensive benchmark of state-of-the-art HAD methods.

РАЗРАБОТКА ПРОГРАММНЫХ СРЕДСТВ МАТЕМАТИЧЕСКОГО ИМИТАЦИОННОГО МОДЕЛИРОВАНИЯ НА ОСНОВЕ КЛИНИЧЕСКИХ ДАННЫХ И ФАНТОМНЫХ ИССЛЕДОВАНИЙ ДЛЯ ОЦЕНКИ ПЕРФУЗИИ ГОЛОВНОГО МОЗГА И ПОВЫШЕНИЯ КАЧЕСТВА ИЗОБРАЖЕНИЙ ПРИ ОФЭКТ/КТ С 99MTC-ГМПАО

Purpose: To develop a software package Virtual examination of brain perfusion by the method of SPECT/CT with 99mTc-HMPAO (Teoxime) and its practical application to study the conditions for achieving the best image quality in clinical studies of patients. Material and methods: The studies were performed using clinical data and the method of computer simulation. Clinical data of single-photon emission computed tomography combined with X-ray computed tomography (SPECT/CT) with 99mTc-hexamethylpropyleneamine oxime (99mTc-Teoxime, produced by DIAMED LLC) of a patient with an ischemic stroke of the right frontal cortex were obtained on a two-detector gamma-camera NM/CT 670 DR GE Discovery (USA) using high-resolution low-energy collimators (LEHR). The measured data were processed using specialized software Q.Brain and Q.Volumetrix MI on a Xeleris 4.0 DR workstation from GE Healthcare (USA) to obtain reconstructed axial tomographic slices. To carry out simulation computer simulation of the procedure of examination of perfusion of GM by the method of SPECT/CT has developed a software package that includes a mathematical Hoffman phantom with the ability to simulate clinical cases of hypoperfusion of different localization and size (Virtual Patient), modeling the collection of “raw” projection data and an image reconstruction program based on the OSEM algorithm (Ordered Subset Expectation Maximization). An important advantage of the mathematical modeling method is the ability to assess the quality of the reconstructed image by calculating the root-mean-square error when compared with a given phantom. In numerical experiments, the dependence of the reconstruction error on the parameters of the OSEM algorithm (on the number of subgroups – subsets, and on the number of iterations) was investigated in order to determine the conditions for achieving the best image quality. A statistical stop criterion was developed and tested. Results: For the first time, a software package was developed and tested that allows us to investigate errors in the reconstruction algorithm, which is a great difficulty when using clinical research methods. A criterion for stopping iterations is proposed when using the OSEM reconstruction algorithm – minimizing the functional deviation of the chi-square function from the target value, while the detector pixels with non-zero values are combined into blocks according to the 2×2 scheme. There is a reliable good correlation between the proposed stop criterion and the minimum of the root-mean-square error of image reconstruction. This makes it possible to introduce this criterion into the clinical practice of using computational tools for reconstructing sections of the SPECT to obtain the best image. The simulation results demonstrated the possibility of reducing the time of data recording, during which the patient must remain motionless, at least twice. Conclusion: The method of computer simulation developed in this work is a practically useful technology that helps optimize the use of SPECT to achieve the best possible results of brain imaging in patients.

Deep expectation-maximization network for unsupervised image segmentation and clustering

Unsupervised learning, such as unsupervised image segmentation and clustering, are fundamental tasks in image representation learning. In this paper, we design a deep expectation-maximization (DEM) network for unsupervised image segmentation and clustering. It is based on the statistical modeling of image in its latent feature space by Gaussian mixture model (GMM), implemented in a novel deep learning framework. Specifically, in the unsupervised setting, we design an auto-encoder network and an EM module over the image latent features, for jointly learning the image latent features and GMM model of the latent features in a single framework. To construct the EM-module, we unfold the iterative operations of EM algorithm and the online EM algorithm in fixed steps to be differentiable network blocks, plugged into the network to estimate the GMM parameters of the image latent features. The proposed network parameters can be end-to-end optimized using losses based on log-likelihood of GMM, entropy of Gaussian component assignment probabilities and image reconstruction error. Extensive experiments confirm that our proposed networks achieve favorable results compared with several state-of-the-art methods in unsupervised image segmentation and clustering.

On the correction of respiratory motion-induced image reconstruction errors in positron-emission tomography-guided radiation therapy

Free breathing (FB) positron emission tomography (PET) images are routinely used in radiotherapy for lung cancer patients. Respiration-induced artifacts in these images compromise treatment response assessment and obstruct clinical implementation of dose painting and PET-guided radiotherapy. The purpose of this study is to develop a blurry image decomposition (BID) method to correct motion-induced image-reconstruction errors in FB-PETs. Assuming a blurry PET is represented as an average of multi-phase PETs. A four-dimensional computed-tomography image is deformably registered from the end-inhalation (EI) phase to other phases. With the registration-derived deformation maps, PETs at other phases can be deformed from a PET at the EI phase. To reconstruct the EI-PET, the difference between the blurry PET and the average of the deformed EI-PETs is minimized using a maximum-likelihood expectation-maximization algorithm. The developed method was evaluated with computational and physical phantoms as well as PET/CT images acquired from three patients. The BID method increased the signal-to-noise ratio from 1.88±1.05 to 10.5±3.3 and universal-quality index from 0.72±0.11 to 1.0 for the computational phantoms, and reduced the motion-induced error from 69.9% to 10.9% in the maximum of activity concentration and from 317.5% to 8.7% in the full width at half maximum of the physical PET-phantom. The BID-based corrections increased the maximum standardized-uptake values by 17.7±15.4% and reduced tumor volumes by 12.5±10.4% on average for the three patients. The proposed image-decomposition method reduces respiration-induced errors in PET images and holds potential to improve the quality of radiotherapy for thoracic and abdominal cancer patients.

Video-rate hyperspectral camera based on a CMOS-compatible random array of Fabry–Pérot filters

Hyperspectral (HS) imaging provides rich spatial and spectral information and extends image inspection beyond human perception. Existing approaches, however, suffer from several drawbacks such as low sensitivity, resolution and/or frame rate, which confines HS cameras to scientific laboratories. Here we develop a video-rate HS camera capable of collecting spectral information on real-world scenes with sensitivities and spatial resolutions comparable with those of a typical RGB camera. Our camera uses compressive sensing, whereby spatial–spectral encoding is achieved with an array of 64 complementary metal–oxide–semiconductor (CMOS)-compatible Fabry–Pérot filters placed onto a monochromatic image sensor. The array affords high optical transmission while minimizing the reconstruction error in subsequent iterative image reconstruction. The experimentally measured sensitivity of 45% for visible light, the spatial resolution of 3 px for 3 dB contrast, and the frame rate of 32.3 fps at VGA resolution meet the requirements for practical use. For further acceleration, we show that AI-based image reconstruction affords operation at 34.4 fps and full high-definition resolution. By enabling practical sensitivity, resolution and frame rate together with compact size and data compression, our HS camera holds great promise for the adoption of HS technology in real-world scenarios, including consumer applications such as smartphones and drones.

A Machine Learning Approach for Atherosclerosis Detection in Murine Models

The early detection of atherosclerosis has been the interest of researchers in order to prevent diseases progression. And, most of cases diagnosed with atherosclerosis in late stages resulted of subjective ways of diagnosis in the healthcare sector. Consequently, in order to help cardiologists diagnose atherosclerosis in early stages with an objective, fast and accurate way, we are proposing a machine learning model based on autoencoders that enables atherosclerosis detection from MRI images of murine subjects. This novel way of automated system consists of applying various image processing techniques on the input MRI images. Then, after training, testing and validation it will be capable of classifying the images as atherosclerotic or not based on a specific threshold for the reconstruction error calculated from the autoencoders' output. An autoencoder is a feed-forward neural network that has its input neurons equal to the output neuron. The classification works by comparing the reconstructed images to the original input images and evaluating loss between them. Since the autoencoder is trained on healthy images, reconstruction error of the healthy images would be low while that of atherosclerotic subjects would be higher. By setting a threshold for the loss, we can classify the images as healthy or atherosclerotic. The pre-processing of these images was made using a Block Matching 3D (BM3D) filter to remove the noise in the images prior to application of a Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance the contrast. The next step included introducing the dataset into the autoencoder to start training on the healthy images, after increasing the images number with augmentation. The results showed a reconstruction loss of 0.018 while using the stacked architecture of the autoencoder and 0.0366 when using the convolutional autoencoder architecture.

Self-Supervision-Augmented Deep Autoencoder for Unsupervised Visual Anomaly Detection.

Deep autoencoder (AE) has demonstrated promising performances in visual anomaly detection (VAD). Learning normal patterns on normal data, deep AE is expected to yield larger reconstruction errors for anomalous samples, which is utilized as the criterion for detecting anomalies. However, this hypothesis cannot be always tenable since the deep AE usually captures the low-level shared features between normal and abnormal data, which leads to similar reconstruction errors for them. To tackle this problem, we propose a self-supervised representation-augmented deep AE for unsupervised VAD, which can enlarge the gap of anomaly scores between normal and abnormal samples by introducing autoencoding transformation (AT). Essentially, AT is introduced to facilitate AE to learn the high-level visual semantic features of normal images by introducing a self-supervision task (transformation reconstruction). In particular, our model inputs the original and transformed images into the encoder for obtaining latent representations; afterward, they are fed to the decoder for reconstructing both the original image and applied transformation. In this way, our model can utilize both image and transformation reconstruction errors to detect anomaly. Extensive experiments indicate that the proposed method outperforms other state-of-the-art methods, which demonstrates the validity and advancement of our model.

A compute-in-memory chip based on resistive random-access memory

Realizing increasingly complex artificial intelligence (AI) functionalities directly on edge devices calls for unprecedented energy efficiency of edge hardware. Compute-in-memory (CIM) based on resistive random-access memory (RRAM)1 promises to meet such demand by storing AI model weights in dense, analogue and non-volatile RRAM devices, and by performing AI computation directly within RRAM, thus eliminating power-hungry data movement between separate compute and memory2–5. Although recent studies have demonstrated in-memory matrix-vector multiplication on fully integrated RRAM-CIM hardware6–17, it remains a goal for a RRAM-CIM chip to simultaneously deliver high energy efficiency, versatility to support diverse models and software-comparable accuracy. Although efficiency, versatility and accuracy are all indispensable for broad adoption of the technology, the inter-related trade-offs among them cannot be addressed by isolated improvements on any single abstraction level of the design. Here, by co-optimizing across all hierarchies of the design from algorithms and architecture to circuits and devices, we present NeuRRAM—a RRAM-based CIM chip that simultaneously delivers versatility in reconfiguring CIM cores for diverse model architectures, energy efficiency that is two-times better than previous state-of-the-art RRAM-CIM chips across various computational bit-precisions, and inference accuracy comparable to software models quantized to four-bit weights across various AI tasks, including accuracy of 99.0 percent on MNIST18 and 85.7 percent on CIFAR-1019 image classification, 84.7-percent accuracy on Google speech command recognition20, and a 70-percent reduction in image-reconstruction error on a Bayesian image-recovery task.

Hiding multiple images into a single image via joint compressive autoencoders

Interest in image hiding has been continually growing. Recently, deep learning-based image hiding approaches improve the hidden capacity significantly. However, the major challenges of the existing methods are that they are difficult to balance between the errors of the modified cover image and those of the recovered secret image. To solve this problem, in this paper, we develop an image hiding algorithm based on a joint compressive autoencoder framework. Further, we propose a novel strategy to enlarge the hidden capacity, i.e., hiding multi-images in one container image. Specifically, our approach provides an extremely high image hidden capacity coupled with small reconstruction errors of the secret image. More importantly, we tackle the trade-off problem of earlier approaches by mapping the image representations in the latent spaces of the joint compressive autoencoder models, leading to both high visual quality of the container image and low reconstruction error the secret image. In an extensive set of experiments, we confirm our proposed approach to outperform several state-of-the-art image hiding methods, yielding high imperceptibility and steganalysis resistance of the container images with high recovery quality of the secret images, while improving the image hidden capacity significantly (four times higher than full-image hiding capacity).

Multi-Sensor Fusion Self-Supervised Deep Odometry and Depth Estimation

This paper presents a new deep visual-inertial odometry and depth estimation framework for improving the accuracy of depth estimation and ego-motion from image sequences and inertial measurement unit (IMU) raw data. The proposed framework predicts ego-motion and depth with absolute scale in a self-supervised manner. We first capture dense features and solve the pose by deep visual odometry (DVO), and then combine the pose estimation pipeline with deep inertial odometry (DIO) by the extended Kalman filter (EKF) method to produce the sparse depth and pose with absolute scale. We then join deep visual-inertial odometry (DeepVIO) with depth estimation by using sparse depth and the pose from DeepVIO pipeline to align the scale of the depth prediction with the triangulated point cloud and reduce image reconstruction error. Specifically, we use the strengths of learning-based visual-inertial odometry (VIO) and depth estimation to build an end-to-end self-supervised learning architecture. We evaluated the new framework on the KITTI datasets and compared it to the previous techniques. We show that our approach improves results for ego-motion estimation and achieves comparable results for depth estimation, especially in the detail area.

Multifrequency Magnetic Induction Tomography for Hemorrhagic Stroke Detection Using an Adaptive Threshold Split Bregman Algorithm

Objective: Magnetic induction tomography (MIT) is viewed as a promising method for brain imaging. Most MIT studies are based on time-difference imaging, which cannot be used for detecting hemorrhagic stroke. However, for many years, multifrequency MIT (mfMIT) has been proposed for the early detection of hemorrhagic stroke, but the focus has primarily been on theoretical analysis and simulations. mfMIT has not been applied in clinical research owing to issues such as inadequate imaging algorithms and mfMIT systems. Methods: A weighted frequency difference adaptive thresholding split Bregman (WFD–ATSB) algorithm is proposed herein and is validated for the first time using a 16-channel mfMIT system developed by our group. Results: The imaging results of a single-hemorrhagic model were evaluated using the total image reconstruction error (as the sum of the position error, area error, and image noise), and the imaging results of a double-hemorrhagic model were evaluated by calculating the correlation coefficient. Simulation results showed that WFD–ATSB can reduce the overall image reconstruction error by 52% and improve the correlation coefficient by 35% compared with that of the traditional algorithm. In phantom experiments, the total image reconstruction error was reduced by 27%, and the correlation coefficient was improved by 15%. Conclusion–Significance: The results confirmed that WFD–ATSB is a more practical and stable algorithm for mfMIT compared with traditional methods, which is promising for the rapid detection of cerebral hemorrhage.

STLS-LADMM-Net: A Deep Network for SAR Autofocus Imaging

Synthetic aperture radar (SAR) can provide high-resolution electromagnetic backscattering images of the illuminated area, playing a significant role in various applications. However, achieving focused SAR images is challenging under sparse sampling and phase error conditions. By exploiting the sparsity or compressibility priors, the state-of-the-art sparsity-driven SAR imaging methods can reconstruct images under the condition of sparse sampling. However, the handcrafted priors used in these methods limit the imaging performance, and the iterative solution schemes reduce the computational efficiency. Besides, the measurement inaccuracy introduced by the phase error also degrades the reconstruction performance of the sparsity-driven imaging methods. To address these issues, a deep network for SAR autofocus imaging is proposed, which alternately performs image reconstruction and phase error estimation. When performing image reconstruction, the sparsity-cognizant total least-square (S-TLS) model is introduced to handle the problem of measurement inaccuracy, contributing to robust reconstruction performance under the condition of phase error. During the implementation of the deep network, a feature transform operator is used to realize data-driven prior knowledge learning and overcome the limitations of handcrafted priors. Moreover, the deep network approach can significantly improve computational efficiency. Experiments on simulated and real data verify the effectiveness and efficiency of the proposed method.

Accurate Quaternion Polar Harmonic Transform for Color Image Analysis

Polar harmonic transforms (PHTs) have been applied in pattern recognition and image analysis. But the current computational framework of PHTs has two main demerits. First, some significant color information may be lost during color image processing in conventional methods because they are based on RGB decomposition or graying. Second, PHTs are influenced by geometric errors and numerical integration errors, which can be seen from image reconstruction errors. This paper presents a novel computational framework of quaternion polar harmonic transforms (QPHTs), namely, accurate QPHTs (AQPHTs). First, to holistically handle color images, quaternion-based PHTs are introduced by using the algebra of quaternions. Second, the Gaussian numerical integration is adopted for geometric and numerical error reduction. When compared with CNNs (convolutional neural networks)-based methods (i.e., VGG16) on the Oxford5K dataset, our AQPHT achieves better performance of scaling invariant representation. Moreover, when evaluated on standard image retrieval benchmarks, our AQPHT using smaller dimension of feature vector achieves comparable results with CNNs-based methods and outperforms the hand craft-based methods by 9.6% w.r.t mAP on the Holidays dataset.

Some aspects of fractional-order circular moments for image analysis

• The importance of optimal parameters in the fractional-order moments is discussed. • The proposed algorithm searches the optimal fractional parameter. • The optimal parameter found is invariant to changes in scale and rotation. • The experimental analysis demonstrates the superiority of the optimal parameter. • The proposed method performs superbly in rotation-invariant object recognition. In this paper, we briefly review the fractional-order circular moments, such as fractional-order Zernike moments, fractional-order Fourier–Mellin moments, fractional-order Legendre–Fourier moments, and fractional-order Chebyshev–Fourier moments, which can characterize, analyze, and manipulate the information contained in an image with minimal redundancy. Also, they depend on an α parameter for better feature extraction. Therefore, we propose a procedure to find the optimal α in terms of image reconstruction error and classification. We validate the search for the best rotation-invariant features using the MNIST and MNIST-R datasets. Finally, we present the study results and conclusions.

Inverse Synthetic Aperture Radar Sparse Imaging Exploiting the Group Dictionary Learning

Sparse imaging relies on sparse representations of the target scenes to be imaged. Predefined dictionaries have long been used to transform radar target scenes into sparse domains, but the performance is limited by the artificially designed or existing transforms, e.g., Fourier transform and wavelet transform, which are not optimal for the target scenes to be sparsified. The dictionary learning (DL) technique has been exploited to obtain sparse transforms optimized jointly with the radar imaging problem. Nevertheless, the DL technique is usually implemented in a manner of patch processing, which ignores the relationship between patches, leading to the omission of some feature information during the learning of the sparse transforms. To capture the feature information of the target scenes more accurately, we adopt image patch group (IPG) instead of patch in DL. The IPG is constructed by the patches with similar structures. DL is performed with respect to each IPG, which is termed as group dictionary learning (GDL). The group oriented sparse representation (GOSR) and target image reconstruction are then jointly optimized by solving a l1 norm minimization problem exploiting GOSR, during which a generalized Gaussian distribution hypothesis of radar image reconstruction error is introduced to make the imaging problem tractable. The imaging results using the real ISAR data show that the GDL-based imaging method outperforms the original DL-based imaging method in both imaging quality and computational speed.

Reducing reconstruction error of classified textural patches by integration of random forests and coupled dictionary nonlinear regressors: with applications to super-resolution of abdominal CT images.

Random forests and dictionary-based statistical regressions have common characteristics, including non-linear mapping and supervised learning. To reduce the reconstruction error of high-resolution images, we integrate random forests and coupled dictionary learning. Textural differences of image blocks are considered by the classification of patches using an Auto-Encoder network. The proposed algorithm partitions an input LR image by 5 × 5 blocks and classifies training patches into six categories. A single random forest regressor is then trained corresponding to each class. The output of an RF is considered as an initial estimate of the HR slice. If a slice's representation is sparse in the Discrete Cosine Transform domain, the initial reconstructed image is further improved by a coupled dictionary. In this study, we applied our method to abdominal CT scans and compared them to conventional and recent researches. We achieved an average improvement of 0.06 (2.37) using the SSIM (PSNR) index compared to the random forest + dictionary learning method. The low standard deviation of the results reveals the stability of the proposed method as well. The proposed algorithm depicts the effectiveness of classifying image patches and individual treatment of each class.