Noise Removal Algorithm Research Articles

Image denoising using block matching and convolutional neural network

Image noise removal (denoising) is a significant preprocessing step in image processing research domain. In past decades, various noise removal algorithms have been developed, and they are categorized into conventional and deep learning denoising methods. Most recently, deep learning-based denoising methods outperform the conventional methods. However, deep learning methods have the limitations of requiring large training sample images, difficult to train the deep convolution neural network (DeCNN), and the majority of the convolutional neural network suffers from the performance saturation. Hence, using combination of block matching (BM) and dilated convolutional neural network (DCNN), a unique method for reducing noise in still images is proposed. In this approach, initially, existing algorithm used to preprocess the image. Then, BM technique uses a sliding window size of a four-by-four block and moves it across the image to select for blocks that are similar in the image. The matching blocks are highly correlated with one another. The matching blocks are then fed to the deep DCNN to remove noise and obtain a better noiseless image. Finally, the noiseless image performances are evaluated using standard metrics. From the experimental results, it is observed that the proposed BM and deep DCNN yielded substantially better results than existing state-of-the-art methods to remove the image’s noise.

The unsupervised pattern recognition for the ECG signal features detection

Smart devices and cloud-based technologies allow continuous non-invasive bio-signal recordings in daily life activities. There are many programs and applications that allow monitoring health condition and sharing that information with other people. However, bio-signals that are recorded in real nonstationary conditions are highly contaminated with noise that depends on various activities. Improper signal processing algorithms may lead to a faulty diagnosis or inaccurate decision-making results. Movement contaminated bio-signals require adaptive filtering and feature extraction algorithms because low frequency trends are mostly unstable, and the noise may cause higher impulses than the signal itself. That is why ordinary ECG signal parameters extraction algorithms fail in real time signal processing. In this research signals were filtered using two different methods: Butterworth filter for the high frequency noise reduction and BEADS algorithm for the low frequency noise removal. A new ECG feature extraction algorithm was proposed that is based on the unsupervised MTEO algorithm together with additional local extremum search. The proposed algorithm was compared with other methods on the MIT-BIH database. The additional comparison was made with Pan-Tompkins and k-TEO algorithms on ECG signals that were recorded in movement. Several examples are presented that show how each algorithm performs during training sessions with various intensity levels. The suggested method performed in linear time complexity that made it sufficient for the real time data processing. The obtained ECG parameter values could be used for diagnostics and fatigue recognition in health monitoring processes.

Data-driven approach for automatic detection of aortic valve opening: B point detection from impedance cardiogram.

Pre-ejection period (PEP), an indicator of sympathetic nervous system activity, is useful in psychophysiology and cardiovascular studies. Accurate PEP measurement is challenging and relies on robust identification of the timing of aortic valve opening, marked as the B point on impedance cardiogram (ICG) signals. The ICG sensitivity to noise and its waveform's morphological variability makes automated B point detection difficult, requiring inefficient and cumbersome expert visual annotation. In this article, we propose a machine learning-based automated algorithm to detect the aortic valve opening for PEP measurement, which is robust against noise and ICG morphological variations. We analyzed over 60 hr of synchronized ECG and ICG records from 189 subjects. A total of 3657 averaged beats were formed using our recently developed ICG noise removal algorithm. Features such as the averaged ICG waveform, its first and second derivatives, as well as high-level morphological and critical hemodynamic parameters were extracted and fed into the regression algorithms to estimate the B point location. The morphological features were extracted from our proposed "variable" physiologically valid search-window related to diverse B point shapes. A subject-wise nested cross-validation procedure was performed for parameter tuning and model assessment. After examining multiple regression models, Adaboost was selected, which demonstrated superior performance and higher robustness to five state-of-the-art algorithms that were evaluated in terms of low mean absolute error of 3.5ms, low median absolute error of 0.0ms, high correlation with experts' estimates (Pearson coefficient=0.9), and low standard deviation of errors of 9.2ms. For reproducibility, an open-source toolbox is provided.

New Splitting Algorithms for Multiplicative Noise Removal Based on Aubert-Aujol Model

New Splitting Algorithms for Multiplicative Noise Removal Based on Aubert-Aujol Model

Arrhythmia detection based on multi-scale fusion of hybrid deep models from single lead ECG recordings: A multicenter dataset study

BackgroundElectrocardiogram automated arrhythmia detection plays a crucial role in the early prevention and diagnosis of cardiovascular diseases. However, previous research relies on noise removal algorithms and extracting solid features from raw ECGs. Besides, existing heartbeat classifiers ignore underlying complementary information of various scales, and intra-patient paradigms often lead to biased results. MethodsWe constructed a novel end-to-end Multi-Scale Convolutional Neural Network-Sequence to Sequence architecture for heartbeat classification to address these issues. We have verified this approach on the clinical data collected by wearable devices and two heterogeneous datasets. ResultsThe proposed model can effectively capture multi-granularity frequency and longitudinal temporal information by fusion representation and sequence learning. The overall F1 score of our approach was achieved at 99.57%, which exceeded the reference pure cascade model by 4.36%.

Critical comparison of background correction algorithms used in chromatography

The objective of the present work was to make a quantitative and critical comparison of a number of drift and noise-removal algorithms, which were proven useful by other researchers, but which had never been compared on an equal basis. To make a rigorous and fair comparison, a data generation tool is developed in this work, which utilizes a library of experimental backgrounds, as well as peak shapes obtained from curve fitting on experimental data. Several different distribution functions are used, such as the log-normal, bi-Gaussian, exponentially convoluted Gaussian, exponentially modified Gaussian and modified Pearson VII distributions. The tool was used to create a set of hybrid (part experimental, part simulated) data, in which the background and all peak profiles and areas are known. This large data set (500 chromatograms) was analysed using seven different drift-correction and five different noise-removal algorithms (35 combinations). Root-mean square errors and absolute errors in peak area were determined and it was shown that in most cases the combination of sparsity-assisted signal smoothing and asymmetrically reweighted penalized least-squares resulted in the smallest errors for relatively low-noise signals. However, for noisier signals the combination of sparsity-assisted signal smoothing and a local minimum value approach to background correction resulted in lower absolute errors in peak area. The performance of correction algorithms was studied as a function of the density and coverage of peaks in the chromatogram, shape of the background signal, and noise levels. The developed data-generation tool is published along with this article, so as to allow similar studies with other simulated data sets and possibly other algorithms. The rigorous assessment of correction algorithms in this work may facilitate further automation of data-analysis workflows.

Thermal Denoising of Cross-Polarized Sentinel-1 Data in Interferometric and Extra Wide Swath Modes

An updated algorithm for removal of thermal noise in Sentinel-1 synthetic aperture radar (SAR) Level-1 Ground Range Detected (GRD) data in cross-polarization is presented. The algorithm is comprised of two steps: correction of the annotated thermal noise magnitude (previously proposed in park2018) and a novel correction of the annotated thermal noise range dependence. The magnitude of the annotated thermal noise is corrected by applying scale and offset coefficients tuned on a few hundred of Sentinel-1 data acquired over surfaces with low backscatter in Interferometric Wide and Extra Wide swath modes in HV and VH polarizations. The values of coefficients for all modes and polarizations for data processed with Instrument Processing Facility (IPF) version 3.1 -3.3 are provided. The range dependence is corrected by minimizing a cost function between the annotated range profiles of thermal noise and antenna pattern gain (APG). An objective validation metric based on comparison of averaged backscatter at inter-swath boundaries is proposed. Validation is performed on hundreds of Sentinel-1 scenes acquired over open ocean, doldrums, deserts and sea ice. It shows that the new algorithm outperforms the standard thermal noise removal algorithm proposed by European Space Agency in almost all cases. Analysis shows that the new algorithm worsens noise correction in cases when the range dependence of the annotated APG does not match with the observed signal, indicating either problems with signal processing on IPF or imprecise annotation of APG.

Mixed noise removal algorithm of EBAPS image under low illumination condition based on FPGA

为了解决单一的中值滤波和高斯滤波算法对低照度图像中同时存在的脉冲噪声和泊松噪声抑制效果不佳、边缘细节保护不足的问题，提出一种基于可编程逻辑门阵列(field programmable gate array，FPGA)的开关中值-高斯融合(open and close mix-median-Gaussian，OCMMG)滤波算法。首先，利用最小四方向差值检测每个像素点的异常程度，根据脉冲噪声判别阈值分配权重，进行第1步滤波处理；然后，利用四方向边缘检测算法提取图像边缘，根据设置的边缘置信度表征值进行第2步滤波处理；最后，用电子轰击有源像素传感器(electron bombarded active pixel sensor，EBAPS)在1×10⁻³ lx照度条件下采集的图像，基于FPGA对其进行实时图像处理。实验结果表明，FPGA处理结果与软件仿真处理结果相符。该算法相比于中值滤波和高斯滤波算法，峰值信噪比分别提高了3.23%和16.34%，结构相似性分别提高了14.66%和33.86%，边缘保持指数分别提高了0.49%和4.21%，能够有效去除EBAPS图像的混合噪声，并满足实时性要求。

Denoising of Impulse Noise using Partition-Supported Median, Interpolation and DWT in Dental X-Ray Images

The impulse noise often damages the human dental X-Ray images, leading to improper dental diagnosis. Hence, impulse noise removal in dental images is essential for a better subjective evaluation of human teeth. The existing denoising methods suffer from less restoration performance and less capacity to handle massive noise levels. This method suggests a novel denoising scheme called "Noise removal using Partition supported Median, Interpolation, and Discrete Wavelet Transform (NRPMID)" to address these issues. To effectively reduce the salt and pepper noise up to a range of 98.3 percent noise corruption, this method is applied over the surface of dental X-ray images based on techniques like mean filter, median filter, Bi-linear interpolation, Bi-Cubic interpolation, Lanczos interpolation, and Discrete Wavelet Transform (DWT). In terms of PSNR, IEF, and other metrics, the proposed noise removal algorithm greatly enhances the quality of dental X-ray images.

A Noise-Removal Algorithm Without Input Parameters Based on Quadtree Isolation for Photon-Counting LiDAR

The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) is the world’s first satellite-borne photon-counting laser altimeter with unprecedented detection performance. Noise removal is an important process applied to raw data and determines the quality of the end product. Assuming that the sparse spatial distribution of noise photons makes them more easily isolated than signal photons, we propose a noise-removal algorithm without input parameters based on quadtree isolation. MATLAS was used to evaluate the performance of our algorithm. We compare our algorithm to the improved density-based spatial clustering of applications with noise (DBSCAN) algorithm. Experimental results show that our algorithm accurately extracts signal photons from raw data and is superior to the improved DBSCAN in accuracy and time efficiency. This novel algorithm makes it possible to efficiently remove noise from photon-counting light detection and ranging (LiDAR) data.

Application of an Automatic Noise or Signal Removal Algorithm Based on Synchrosqueezed Continuous Wavelet Transform of Passive Surface Wave Imaging: A Case Study in Sichuan, China

Passive surface wave imaging based on noise cross-correlation has been a research hotspot in recent years. However, because randomness of noise is difficult to achieve in reality, prominent noise sources will inevitably affect the dispersion measurement. Additionally, in order to recover high-fidelity surface waves, the time series input during cross-correlation calculation is usually very long, which greatly limits the efficiency of passive surface wave imaging. With an automatic noise or signal removal algorithm based on synchrosqueezed continuous wavelet transform (SS-CWT), these problems can be alleviated. We applied this method to 1-h passive datasets acquired in Sichuan province, China; separated the prominent noise events in the raw field data, and enhanced the cross-correlation reconstructed surface waves, effectively improving the accuracy of the dispersion measurement. Then, using the conventional surface wave inversion method, the shear wave velocity profile of the underground structure in this area was obtained.

Noise reduction in speckle interferometry fringe pattern using adaptive Kalman filter

Fringe pattern denoising is of prime importance in phase demodulation, especially for a single fringe pattern in speckle interferometry. A fringe speckle noise removal algorithm using the Kalman filter is proposed. The conventional linear Kalman filter is implemented with a fixed value of the process and measurement noise covariances; the adaptive Kalman filter is implemented with the process, and measurement noise covariances are estimated in an adaptive manner. The fringe denoising is performed sequentially in a rowwise and columnwise manner. The simulated and experimental results demonstrate the practical applicability of the proposed method, and its performance is appraised using metrics such as the quality index, peak signal-to-noise-ratio, and speckle index.

Retrieving building height in urban areas using ICESat-2 photon-counting LiDAR data

Building height is a pivotal factor in studying urban form and understanding the impacts of the vertical characteristics of urban areas on the environment. The Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) employ photon-counting LiDAR to collect Earth’s surface elevation data globally. However, method at ground and building photon classification and subsequently building height extraction from photon-counting LiDAR data is still missing. This study aimed to develop a methodological framework to retrieve building height from ICESat-2 data under various data acquisition conditions stratified by day/night condition, strong/weak beam. First, the noise removal algorithm adopted by ATL03 product is utilized to filter out noise photons and retain signal photons in the raw dataset. Second, the random sample consensus (RANSAC) algorithm and statistical method are adopted to classify the signal photons into ground and building photons according to their characteristics and spatial neighborhood relationship. The building height is then computed by the elevation of building and ground photons. Finally, estimated heights are evaluated using the reference building height derived from terrestrial laser scanning (TLS) data. The results indicate that the proposed methodological framework can effectively identify building and ground photons under different building types. Strong consistence between estimated and reference building height is obtained by quantifying the estimation precision of building height under different data acquisition conditions, root mean square error (RMSE) ranging between 0.35 m and 0.45 m, indicating that the ATL03 data can be utilized to derive the building height in urban areas for all acquisition times and beam intensity. Further analysis demonstrates that strong/weak beam significantly influences the building height estimation compared to day/night condition. Overall, this study provides a method for estimating building height in urban areas using ICESat-2 data, and the findings approve the strong capability of ICESat-2 data in estimating building height.

Deep Learning Based Filtering Algorithm for Noise Removal in Underwater Images

Under-water sensing and image processing play major roles in oceanic scientific studies. One of the related challenges is that the absorption and scattering of light in underwater settings degrades the quality of the imaging. The major drawbacks of underwater imaging are color distortion, low contrast, and loss of detail (especially edge information). The paper proposes a method to address these issues by de-noising and increasing the resolution of the image using a model network trained on similar data. The network extracts frames from a video and filters them with a trigonometric–Gaussian filter to eliminate the noise in the image. It then applies contrast limited adaptive histogram equalization (CLAHE) to improvise the image contrast, and finally enhances the image resolution. Experimental results show that the proposed method could effectively produce enhanced images from degraded underwater images.

Speckle noise removal based on structural convolutional neural networks with feature fusion for medical image

Application of convolutional neural networks (CNNs) for image additive white Gaussian noise (AWGN) removal has attracted considerable attentions with the rapid development of deep learning in recent years. However, the work of image multiplicative speckle noise removal is rarely done. Moreover, most of the existing speckle noise removal algorithms are based on traditional methods with human priori knowledge, which means that the parameters of the algorithms need to be set manually. Nowadays, deep learning methods show clear advantages on image feature extraction. Multiplicative speckle noise is very common in real life images, especially in medical images. In this paper, a novel neural network structure is proposed to recover noisy images with speckle noise. Our proposed method mainly consists of three subnetworks. One network is rough clean image estimate subnetwork. Another is subnetwork of noise estimation. The last one is an information fusion network based on U-Net and several convolutional layers. Different from the existing speckle denoising model based on the statistics of images, the proposed network model can handle speckle denoising of different noise levels with an end-to-end trainable model. Extensive experimental results on several test datasets clearly demonstrate the superior performance of our proposed network over state-of-the-arts in terms of quantitative metrics and visual quality.

Noise Removal Algorithm for Out-of-focus Blurred Images Based on Bilateral Filtering

The feature resolution of traditional methods for fuzzy image denoising is low, for the sake of improve the strepitus removal and investigation ability of defocused blurred night images, a strepitus removal algorithm based on bilateral filtering is suggested. The method include the following steps of: Building an out-of-focus blurred night scene image acquisition model with grid block feature matching of the out-of-focus blurred night scene image; Carrying out information enhancement processing of the out-of-focus blurred night scene image by adopting a high-resolution image detail feature enhancement technology; Collecting edge contour feature quantity of the out-of-focus blurred night scene image; Carrying out grid block feature matching design of the out-of-focus blurred night scene image by adopting a bilateral filtering information reconstruction technology; And building the gray-level histogram information location model of the out-of-focus blurred night scene image. Fuzzy pixel information fusion investigation method is used to collect gray features of defocused blurred night images. According to the feature collection results, bilateral filtering algorithm is used to automatically optimize the strepitus removal of defocused blurred night images. The simulation results show that the out-of-focus blurred night scene image using this method for machine learning has better strepitus removal performance, shorter time cost and higher export peak signal-to-strepitus proportion.

Limited-angle weighted ultrasonic back-projection imaging with ART algorithm imaging

Ultrasonic through-transmission data processed using the back-projection algorithm offers depth and lateral information about a defect beyond the capabilities of current through-transmission techniques. This technique was trialled on a carbon steel block containing side-drilled holes. Imaging artefacts can arise from the use of the backprojection algorithm, due to applying a weighting of one to each pixel, irrespective of how much of the pixel is intersected by the beam. Noise can also occur within the image where there are few intersections of the pixels made. This is seen at the edges of the image. In this paper, a novel back-projection technique utilises the weighting of pixels, dependent on the normalised weight of the beam that intersects them, to reduce any artefacts that occurred previously due to the backprojection algorithm. This paper also explores the use of the algebraic reconstruction technique (ART) algorithm for noise removal, thus increasing the sharpness of the defect.

Leaf area index retrieval with ICESat-2 photon counting LiDAR

This is the first study to inverse leaf area index (LAI) with photon counting LiDAR (i.e., Ice, Cloud, and Land Elevation Satellite-2/Advanced Topographic Laser Altimeter System, ICESat-2/ATLAS). A new framework composed of three major steps was applied: noise removal algorithm, photon classification algorithm, and the LAI estimation model. To evaluate the feasibility and accuracy of ICESat-2 derived LAI, the MODIS (Moderate Resolution Imaging Spectroradiometer) LAI product at the spatial resolution of 500-m over two sites (Amazon rainforest and Daxing’an mountains forests) were selected for comparison purpose. The results show satisfactory agreements (Amazon: R=0.693, RMSE=2.545; Daxing’an: R=0.626, RMSE=1.893; Two sites together: R=0.667, RMSE=2.433) between MODIS LAI and ICESat-2 derived LAI. Moreover, we also selected the LAI derived from Sentinel-2 to further validate the ICESat-2 derived LAI, and got better results than MODIS LAI (Amazon: R=0.752, RMSE=2.329; Daxing’an: R=0.704, RMSE=1.724). Our study found that: (1) The new applied framework can effectively inverse LAI with ICESat-2. (2) The ICESat-2 derived LAI are reliable. (3) ICESat-2 exists many advantages over MODIS in terms of LAI estimation, for example, it not only can alleviate the issue of saturation of MODIS when LAI is high, but also can mitigate the problem of poor inversion effect of MODIS where the vegetation types are diversity.

Performance Comparison of SAR Image Speckle Noise Removal Algorithms

Performance Comparison of SAR Image Speckle Noise Removal Algorithms

Remote sensing image striping noise algorithm based on region-weighted double sparse constraint

Infrared remote sensing image has been used in many fields such as land cover crop monitoring, climate monitoring, military strike and early warning. In spaceborne remote sensing imaging system, the stripe is a common noise, seriously affects the researchers’ study of the ground scene. In recent years, the framework that researchers based on different principles use a variety of methods to remove fringe noise in infrared remote sensing images is proposed, such as filtering based method, the method based on statistics, however, most current algorithm still exist the following two problems: some structure or texture was easy to be mistaken as the stripe noise and removed, lost the original information of the image; in some bright or dark areas, it is easy to create artificial traces, adding redundant information and making the denoising result look unnatural. In view of the above two problems, this paper will analyse the causes of the problems and propose a stripe noise removal algorithm based on the region-weighted doublesparse constrained unidirectional variational model, which can effectively reduce the above two phenomena.