Levels Of Gaussian Noise Research Articles

Surface-Related Multiples Elimination for Waterborne GPR Data

Ground-penetrating radar (GPR) is a well-respected, effective, and efficient geophysical technique. However, for underwater engineering detection and underwater archaeology, the measured B-scan profiles typically contain surface-related multiple waves, which can reduce the signal to noise ratio and interfere with the interpretation of results. SRME is a feedback iteration method based on wave equation, which is frequently utilized in marine seismic explorations but very rarely in GPR underwater engineering detection. To fill this gap, we applied SRME to suppress multiples that appear in GPR underwater images. When we compared the effectiveness of the underwater horizontal layered model and the underwater undulating interface model, we found a high match rate between the predicted and the real-world multiples. In addition, the addition of the Gaussian random noise level with a 4% maximum amplitude to the B-scan profile of the horizontal stratified model yielded satisfactory multiple suppression results. Finally, we applied this method to the B-scan GPR section of actual underwater archaeological images to achieve multiple suppression, which can more effectively weaken and inhibit the surface-related multiples. Both numerical simulations and actual field data show that the SRME method is highly suitable for interpreting waterborne GPR data, and more accurate interpretation can be obtained from the GPR profile after multiples suppression.

Self-supervision for Tabular Data by Learning to Predict Additive Homoskedastic Gaussian Noise as Pretext

The lack of scalability of data annotation translates to the need to decrease dependency on labels. Self-supervision offers a solution with data training themselves. However, it has received relatively less attention on tabular data, data that drive a large proportion of business and application domains. This work, which we name the Statistical Self-Supervisor (SSS), proposes a method for self-supervision on tabular data by defining a continuous perturbation as pretext. It enables a neural network to learn representations by learning to predict the level of additive isotropic Gaussian noise added to inputs. The choice of the pretext transformation is motivated by intrinsic characteristics of a neural network fundamentally performing linear fits under the widely adopted assumption of Gaussianity in its fitting error and the preservation of locality of a data example on the data manifold in the presence of small random perturbations. The transform condenses information in the generated representations, making them better employable for further task-specific prediction as evidenced by performance improvement of the downstream classifier. To evaluate the persistence of performance under low-annotation settings, SSS is evaluated against different levels of label availability to the downstream classifier (1% to 100%) and benchmarked against self- and semi-supervised methods. At the most label-constrained, 1% setting, we report a maximum increase of at least 2.5% against the next-best semi-supervised competing method. We report an increase of more than 1.5% against self-supervised state of the art. Ablation studies also reveal that increasing label availability from 0% to 1% results in a maximum increase of up to 50% on either of the five performance metrics and up to 15% thereafter, indicating diminishing returns in additional annotation.

Image Denoising by Deep Convolution Based on Sparse Representation

Noise filtering is a crucial task in digital image processing, performing the function of preprocessing. In this paper, we propose an algorithm that employs deep convolution and soft thresholding iterative algorithms to extract and learn the features of noisy images. The extracted features are acquired through prior and sparse representation theory for image reconstruction. Effective separation of the image and noise is achieved using an end-to-end network of dilated convolution and fully connected layers. Several experiments were performed on public images subject to various levels of Gaussian noise, in order to evaluate the effectiveness of the proposed approach. The results indicated that our algorithm achieved a high peak signal-to-noise ratio (PSNR) and significantly improved the visual effects of the images. Our study supports the effectiveness of our approach and substantiates its potential to be applied to a broad spectrum of image processing tasks.

Image Denoising using Mean Filter

The first step in any image processing process is to remove noise from an image, and this is the aim of every de-noising technique. The noise removal technique should be used carefully to avoid introducing artifacts that could cause the image to become blurry. In this paper, five levels of Gaussian noise and Salat & Peppers are applied to original images to add noise (σ=5, σ=10, σ=15, σ=20, σ=25 σ=50) and also combination Gaussian and Salt & peepers are used with noise level 25. Removing noising is filtered in mean filter. Performance evaluation of the mean filter with three methods of noise is provided. After testing all the approaches with Gaussian noise σ =15, the results revealed that Lena's image was superior to the others. The other photos, however, performed better for the Gaussian levels 20 and 25 than the Lena image. With all types of noise, the mean filter produced good results.

De-Noising of Photoacoustic Microscopy Images by Attentive Generative Adversarial Network.

As a hybrid imaging technology, photoacoustic microscopy (PAM) imaging suffers from noise due to the maximum permissible exposure of laser intensity, attenuation of ultrasound in the tissue, and the inherent noise of the transducer. De-noising is an image processing method to reduce noise, and PAM image quality can be recovered. However, previous de-noising techniques usually heavily rely on manually selected parameters, resulting in unsatisfactory and slow de-noising performance for different noisy images, which greatly hinders practical and clinical applications. In this work, we propose a deep learning-based method to remove noise from PAM images without manual selection of settings for different noisy images. An attention enhanced generative adversarial network is used to extract image features and adaptively remove various levels of Gaussian, Poisson, and Rayleigh noise. The proposed method is demonstrated on both synthetic and real datasets, including phantom (leaf veins) and in vivo (mouse ear blood vessels and zebrafish pigment) experiments. In the in vivo experiments using synthetic datasets, our method achieves the improvement of 6.53 dB and 0.26 in peak signal-to-noise ratio and structural similarity metrics, respectively. The results show that compared with previous PAM de-noising methods, our method exhibits good performance in recovering images qualitatively and quantitatively. In addition, the de-noising processing speed of 0.016 s is achieved for an image with 256×256 pixels, which has the potential for real-time applications. Our approach is effective and practical for the de-noising of PAM images.

Defed: An Edge-Feature-Enhanced Image Denoised Network Against Adversarial Attacks for Secure Internet of Things

With the prosperous development of Internet of Things (IoT), IoT devices have been deployed in various applications, which generates large volume of image data to trace and record the users’ behaviors, resulting in better IoT services. To accurately analyze these huge data to further improve users’ experience on IoT services, deep neural networks (DNNs) are gaining more attention and have become increasingly popular. However, recent studies have shown that DNN models are vulnerable to adversarial attacks, which leads to the risk of applications in practice. Previous works are devoted to extract invariant features from the content circled by edges in images, while such features can not efficiently deal with the adversarial effect. In this work, we first study this problem from a new angle by exploring the edge feature information, which is intractable to be influenced by adversarial attacks demonstrated by our empirical analysis. Based on this, we propose a novel edge feature enhanced defense approach called Defed which incorporates edge feature information into denoised network to defend against various adversarial attacks in image area. For training phase, we only add benign images as the input and exert Gaussian noise to substitute the adversarial attacks to mitigate the dependency of models on specific adversarial attacks. For inference, we design a combination of multiple Defeds trained by different Gaussian noise levels and deploy confidence intervals to judge whether an image is adversarial or not. Experiments over real-world datasets on image classification demonstrate the efficacy and superiority compare to the state-of-the-art defense approaches.

Biocomputing Model Using Tripartite Synapses Provides Reliable Neuronal Logic Gating With Spike Pattern Diversity.

Biocomputing technologies exploit biological communication mechanisms involving cell-cell signal propagation to perform computations. Researchers recently worked toward realising logic gates made by neurons to develop novel devices such as organic neuroprostheses or brain implants made by cells, herein termed living implants. Several challenges arise from this approach, mainly associated with the stochastic nature and noise of neuronal communication. Since astrocytes play a crucial role in the regulation of neurons activity, there is a possibility whereby astrocytes can be engineered to control synapses favouring reliable biocomputing. This work proposes a mathematical model of neuronal logic gates involving neurons and astrocytes, realising OR and AND gating. We use a shallow coupling of both the Izhikevich and Postnov models to characterise gating responses with spike pattern variability and astrocyte synaptic regulation. Logic operation error ratio and accuracy assess the AND and OR gates' performances at different synaptic Gaussian noise levels. Our results demonstrate that the astrocyte regulating activity can effectively be used as a denoising mechanism, paving the way for highly reliable biocomputing implementations.

CTIS-GAN: computed tomography imaging spectrometry based on a generative adversarial network.

Computed tomography imaging spectrometry (CTIS) is a snapshot hyperspectral imaging technique that can obtain a three-dimensional (2D spatial+1D spectral) data cube of the scene captured within a single exposure. The CTIS inversion problem is typically highly ill-posed and is usually solved by time-consuming iterative algorithms. This work aims to take the full advantage of the recent advances in deep-learning algorithms to dramatically reduce the computational cost. For this purpose, a generative adversarial network is developed and integrated with self-attention, which cleverly exploits the clearly utilizable features of zero-order diffraction of CTIS. The proposed network is able to reconstruct a CTIS data cube (containing 31 spectral bands) in milliseconds with a higher quality than traditional methods and the state-of-the-art (SOTA). Simulation studies based on real image data sets confirmed the robustness and efficiency of the method. In numerical experiments with 1000 samples, the average reconstruction time for a single data cube was ∼16m s. The robustness of the method against noise is also confirmed by numerical experiments with different levels of Gaussian noise. The CTIS generative adversarial network framework can be easily extended to solve CTIS problems with larger spatial and spectral dimensions, or migrated to other compressed spectral imaging modalities.

Synergy between Semantic Segmentation and Image Denoising via Alternate Boosting

The capability of image semantic segmentation may be deteriorated due to the noisy input image, where image denoising prior to segmentation may help. Both image denoising and semantic segmentation have been developed significantly with the advance of deep learning. In this work, we are interested in the synergy between these two tasks by using a holistic deep model. We observe that not only denoising helps combat the drop of segmentation accuracy due to the noisy input, but also pixel-wise semantic information boosts the capability of denoising. We then propose a boosting network to perform denoising and segmentation alternately. The proposed network is composed of multiple segmentation and denoising blocks (SDBs), each of which estimates a semantic map and then uses the map to regularize denoising. Experimental results show that the denoised image quality is improved substantially and the segmentation accuracy is improved to close to that on clean images, and segmentation and denoising are both boosted as the number of SDBs increases. On the Cityscapes dataset, using three SDBs improves the denoising quality to 34.42 dB in PSNR, and the segmentation accuracy to 66.5 in mIoU, when the additive white Gaussian noise level is 50.

An application of deep dual convolutional neural network for enhanced medical image denoising.

This work investigates the medical image denoising (MID) application of the dual denoising network (DudeNet) model for chest X-ray (CXR). The DudeNet model comprises four components: a feature extraction block with a sparse mechanism, an enhancement block, a compression block, and a reconstruction block. The developed model uses residual learning to boost denoising performance and batch normalization to accelerate the training process. The name proposed for this model is dual convolutional medical image-enhanced denoising network (DCMIEDNet). The peak signal-to-noise ratio (PSNR) and structure similarity index measurement (SSIM) are used to assess the MID performance for five different additive white Gaussian noise (AWGN) levels of σ = 15, 25, 40, 50, and 60 in CXR images. Presented investigations revealed that the PSNR and SSIM offered by DCMIEDNet are better than several popular state-of-the-art models such as block matching and 3D filtering, denoising convolutional neural network, and feature-guided denoising convolutional neural network. In addition, it is also superior to the recently reported MID models like deep convolutional neural network with residual learning, real-valued medical image denoising network, and complex-valued medical image denoising network. Therefore, based on the presented experiments, it is concluded that applying the DudeNet methodology for DCMIEDNet promises to be quite helpful for physicians.

A Feature-Map-Based Method for Explaining the Performance Degradation of Ship Detection Networks

The unknowability of the inner workings limits the magnitude of performance improvement of ship target detection networks in synthetic aperture radar (SAR) images under Gaussian noise. However, none of the existing interpretation methods explain the phenomenon of network changes under noise. The feature map can visually reflect the changes in image delivery in the network, and some metrics can quantitatively characterize the degree of network performance degradation in a noise environment. So in this paper, we propose a comprehensive analysis method that integrates texture and brightness features of the internal feature map of the network to clarify the change process of target features under Gaussian noise. First, we analyzed the degradation of three target detection networks under different levels of Gaussian noise; then, the feature maps of four convolution layers were sampled and visualized for qualitative analysis; finally, the texture and brightness features were extracted for quantitative characterization of the feature amount changes. We experimentally validated the method on publicly available SSDD radar datasets. The networks were extremely sensitive to Gaussian noise, and the mean Average Precision decreased by up to 96.3%. The angular second moment and entropy texture feature values of the feature map could drop and rise 59.10% and 97.81%, respectively, while the brightness value could increase up to 100.92%. This indicates that noise changes the structure of feature maps and reduces the amount of effective information.

IMPROVING ACCURACY AND REPEATABILITY OF T2 MAPPING IN THE OAI DATA THROUGH EXTEND PHASE GRAPH MODELING

IMPROVING ACCURACY AND REPEATABILITY OF T2 MAPPING IN THE OAI DATA THROUGH EXTEND PHASE GRAPH MODELING

Deep Learning Models for Stable Gait Prediction Applied to Exoskeleton Reference Trajectories for Children With Cerebral Palsy

Gait trajectory prediction models have several applications in exoskeleton control; they can be used as feed-forward input to low-level controllers and to generate reference/target trajectories for position-controlled exoskeletons. In our study, we implement four deep learning models (LSTM, FCN, CNN and Transformer) that perform one-step-ahead gait trajectory prediction after training on gait patterns of typically developing children. We propose a methodology that optimises for stability in long-term forecasts, and evaluate the performance of the models on typically developing (TD) and Cerebral Palsy (CP) gait during recursive prediction of 200 time-steps in the future (which may lead to propagation of errors) and in the presence of varying levels of Gaussian noise (1%-5%). Results on TD gait show that the FCN and Transformer, with mean absolute errors (MAEs) for one-step-ahead predictions between 1.17°-1.63°, are the most suitable for the intended application. We also proposed an approach for generating adaptive trajectories that can be used as reference trajectories for position-controlled exoskeletons. Gait patterns from children with Cerebral Palsy were fed into gait trajectory prediction models trained on typically developing gait only, to generate corrective patterns. Preliminary results show that the gait patterns of typically developing children were introduced onto the generated trajectories.

Accuracy of Rockfall Volume Reconstruction from Point Cloud Data—Evaluating the Influences of Data Quality and Filtering

Rockfall processes are now commonly studied through monitoring campaigns using repeat lidar scanning. Accordingly, several recent studies have evaluated how the temporal resolution of data collection and various data-processing decisions can influence the apparent rockfall volumes estimated using typical rockfall database creation workflows. However, there is a lack of studies that consider how data quality and associated data-processing decisions influence rockfall volume estimation. In this work, we perform a series of tests based on an existing reference rockfall database from the Front Range of Colorado, USA, to isolate the influences of data resolution (point spacing), individual point precision, and the filter threshold applied to change results, on the volume estimates obtained for rockfalls. While the effects of individual point precision were found to be limited for typical levels of gaussian noise (standard deviation per coordinate direction ≤ 0.02 m), data resolution and change filter threshold were found to have systematic impacts on volume estimates, with the volume estimates for the smallest rockfalls decreasing substantially with increases in point spacing and change filter threshold. Because these factors disproportionately impact volume estimates for smaller rockfalls, when these factors change, the slope of the apparent power law that describes the relative frequency-volume distribution of rockfalls changes. Evidence is presented that suggests that this phenomenon can explain discrepancies between power law slopes presented in the literature based on studies focused on different scales of rockfall activity. Overall, this study demonstrates the impacts of raw data attributes on rockfall volume estimation and presents an additional effect that tends to bias rockfall frequency–magnitude power law relationships towards underestimation of the relative prevalence of small rockfalls.

Effects of Image Quality on the Accuracy Human Pose Estimation and Detection of Eye Lid Opening/Closing Using Openpose and DLib.

The application of computer models in continuous patient activity monitoring using video cameras is complicated by the capture of images of varying qualities due to poor lighting conditions and lower image resolutions. Insufficient literature has assessed the effects of image resolution, color depth, noise level, and low light on the inference of eye opening and closing and body landmarks from digital images. This study systematically assessed the effects of varying image resolutions (from 100 × 100 pixels to 20 × 20 pixels at an interval of 10 pixels), lighting conditions (from 42 to 2 lux with an interval of 2 lux), color-depths (from 16.7 M colors to 8 M, 1 M, 512 K, 216 K, 64 K, 8 K, 1 K, 729, 512, 343, 216, 125, 64, 27, and 8 colors), and noise levels on the accuracy and model performance in eye dimension estimation and body keypoint localization using the Dlib library and OpenPose with images from the Closed Eyes in the Wild and the COCO datasets, as well as photographs of the face captured at different light intensities. The model accuracy and rate of model failure remained acceptable at an image resolution of 60 × 60 pixels, a color depth of 343 colors, a light intensity of 14 lux, and a Gaussian noise level of 4% (i.e., 4% of pixels replaced by Gaussian noise). The Dlib and OpenPose models failed to detect eye dimensions and body keypoints only at low image resolutions, lighting conditions, and color depths. Our established baseline threshold values will be useful for future work in the application of computer vision in continuous patient monitoring.

A MoS2 Hafnium Oxide Based Ferroelectric Encoder for Temporal-Efficient Spiking Neural Network.

Spiking neural network (SNN), where the information is evaluated recurrently through spikes, has manifested significant promises to minimize the energy expenditure in data-intensive machine learning and artificial intelligence. Among these applications, the artificial neural encoders are essential to convert the external stimuli to a spiking format that can be subsequently fed to the neural network. Here, a molybdenum disulfide (MoS2 ) hafnium oxide-based ferroelectric encoder is demonstrated for temporal-efficient information processing in SNN. The fast domain switching attribute associated with the polycrystalline nature of hafnium oxide-based ferroelectric material is exploited for spike encoding, rendering it suitable for realizing biomimetic encoders. Accordingly, a high-performance ferroelectric encoder is achieved, featuring a superior switching efficiency, negligible charge trapping effect, and robust ferroelectric response, which successfully enable a broad dynamic range. Furthermore, an SNN is simulated to verify the precision of the encoded information, in which an average inference accuracy of 95.14% can be achieved, using the Modified National Insitute of Standards and Technology(MNIST) dataset for digit classification. Moreover, this ferroelectric encoder manifests prominent resilience against noise injection with an overall prediction accuracy of 94.73% under various Gaussian noise levels, showing practical promises to reduce the computational load for the neural network.

Joint synchronization and parameter estimation in OFDM signaling

AbstractChallenges in cognitive radio and tactical communications include recognizing anonymously received signals and estimating parameters in a blind or semiblind manner. In this paper, we examine this issue for orthogonal frequency division multiplexing (OFDM) signaling. There are several parameters in OFDM signaling, and the blind receiver must extract and consider the synchronization issue. We assume that the blind receiver is aware of modulation type, OFDM, and not aware of chip duration and the length of cyclic prefix. First, we present new criteria based on kurtosis to estimate these parameters and compare their performance at different levels of additive white Gaussian noise with methods based on correlation, kurtosis, maximum likelihood, and matched filter. Then, we perform synchronization and estimate the start time based on these criteria and several new criteria in two steps: fine and coarse synchronization. Finally, in a more practical setup, we present the idea of jointly estimating the mentioned parameters and the signal start time as coarse synchronization. We compare different criteria and show that one of the proposed criteria has the highest efficiency.

Research on extraction method of tunnel magnetic resonance detection signal based on collaborative filtering

Magnetic resonance detection of tunnel as a non-invasive, direct and quantitative geophysical method for detecting groundwater has attracted much attention in the research of tunnel water hazard early warning. In view of the complex environment where the magnetic resonance detection signal is only nanovolt and the tunnel space, the peak noise and environmental noise are much greater than those of ground magnetic resonance detection. In this paper, we propose a peak noise suppression method based on collaborative filtering to suppress the peak noise of tunnel magnetic resonance rotation detection. In this method, the co-filtering parameters are calculated by using the data without peak noise, and then the peak noise suppression of rotation detection is realized. Through simulation, this method can effectively suppress the peak noise in the tunnel rotation detection magnetic resonance signal and improve the signal quality without changing the relaxation and attenuation characteristics of the tunnel magnetic resonance detection signal. The noise reduction effect of peak noise suppression method based on collaborative filtering is compared and analyzed when the amplitude and quantity of peak noise are different. The influence of this method on the peak noise suppression effect under different Gaussian noise levels is discussed. It is concluded that collaborative filtering can suppress the peak and part of Gaussian noise well without losing the effective magnetic resonance signal.

The effect of Gaussian noise on pneumonia detection on chest radiographs, using convolutional neural networks

IntroductionChest X-rays (CXR) with under-exposure increase image noise and this may affect convolutional neural network (CNN) performance. This study aimed to train and validate CNNs for classifying pneumonia on CXR as normal or pneumonia acquired at different image noise levels. MethodsThe study used the curated and publicly available “Chest X-Ray Pneumonia” dataset of 5856 AP CXR classified into 1583 normal, 4273 viral and bacterial pneumonia cases. Gaussian noise with zero mean was added to the images, at 5 image noise variance levels, corresponding to decreasing exposure. Each noise-level dataset was split into 80% for training, 10% for validation, and 10% for test data and then classified using custom trained sequential CNN architecture. Six classification tasks were developed for five Gaussian noise levels and the original dataset. Sensitivity, specificity, predictive values and accuracy were used as evaluation performance metrics. ResultsCNN evaluation on the different datasets revealed no performance drop from the original dataset to the five datasets with different noise levels. Sensitivity, specificity and accuracy for the normal datasets were 98.7%, 76.1% and 90.2%. For the five Gaussian noise levels the sensitivity, specificity and accuracy ranged from 96.9% to 98.2%, 94.4%–98.7% and 96.8%–97.6%, respectively. A heat map was used for visual explanation of the CNNs. ConclusionThe CNNs sensitivity maintained, and the specificity increased in distinguishing between normal and pneumonia CXR with the introduction of image noise. Implications for practiceNo performance drops of CNNs in distinguishing cases with and without pneumonia CXR with different Gaussian noise levels was observed. This has potential for decreasing radiation dose to patients or maintaining exposure parameters for patients that require additional radiographs.

Field programmable gate arrays implementation of different standard deviation estimation techniques

Additive white Gaussian noise level estimation has found its application in many fields such as biomedical signal processing, communication system, and image processing. Many methods have been proposed with different output accuracy, system complexity, power consumption, and speed. In this paper, three of the most well-known and largely used algorithms (median based, root mean square (RMS) based, and P84 based methods) have been implemented and investigated in a full comparison between them to find their advantage and disadvantage, and the suitability of each method for a specific application. The three designs are created using Xilinx system generator (XSG) and implemented on Xilinx field programmable gate arrays (FPGA) development board with Zynq series "XC7Z020-1CLG484", to evaluate the design's performance and the results are discussed in the paper.