Multiscale Filtering Research Articles

Contrast enhancement method in aero thermal radiation images based on cyclic multi-scale illumination self-similarity and gradient perception regularization.

In aerospace, the effects of thermal radiation severely affect the imaging quality of infrared (IR) detectors, which blur the scene information. Existing methods can effectively remove the intensity bias caused by the thermal radiation effect, but they have limitations in the ability of enhancing contrast and correcting local dense intensity or global dense intensity. To address the limitations, we propose a contrast enhancement method based on cyclic multi-scale illumination self-similarity and gradient perception regularization solver (CMIS-GPR). First, we conceive to correct for intensity bias by amplifying gradient. Specifically, we propose a gradient perception regularization (GPR) solver to correct intensity bias by directly decomposing degraded image into a pair of high contrast images, which do not contain intensity bias and exhibit inverted intensity directions. However, we find that the GPR fails for dense intensity area due to small gradient of the scene. Second, to cope with the cases of dense intensity, we regard the dense intensity bias as the sum of multiple slight intensity bias. Then, we construct a cyclic multi-scale illumination self-similarity (CMIS) model by using multi-scale Gaussian filters and structural similarity prior to removing the dense intensity layer by layer. The result acts as coarse correction for GPR, which does not need to be overly concerned with whether the result has intensity residuals or not. Finally, the coarse corrected result is input to the GPR module to further correct residual intensity bias by enhancing contrast. Extensive experiments in real and simulated data have demonstrated the superiority of the proposed method.

Infrared-visible image fusion method based on multi-scale shearing Co-occurrence filter

Infrared-visible image fusion aims to achieve salient targets and abundant texture details. However, most of fusion methods based on multi-scale decomposition fail to effectively preserve detailed and texture information. To avoid these issues, a novel fusion method is proposed by means of multi-scale shearing Co-occurrence filter (MSCOF). This multi-scale decomposition tool possesses the flexible multi-direction and shift-invariant properties of the Shearlet, as well as the smoothing advantages of the Co-occurrence filter (COF). In this study, we obtain approximate and detail subbands by performing MSCOF decomposition on the original images. A weighted fusion rule based on the saliency intensity value (SIV) is performed to combine approximate subbands. Notably, COF is employed on the saliency intensity maps to better preserve edge information of the original image on the fused approximate subbands. Meanwhile, the improved local gradient energy (ILGE) method is utilized to fuse the detail subbands. The results of experimental analysis fully prove that our method is reliable and feasible, and that the obtained objective evaluation metrics are superior to those of some typical methods.

Multimodal Medical Image Fusion Utilizing Two-scale Image Decomposition via Saliency Detection.

Modern medical imaging modalities used by clinicians have many applications in the diagnosis of complicated diseases. These imaging technologies reveal the internal anatomy and physiology of the body. The fundamental idea behind medical image fusion is to increase the image's global and local contrast, enhance the visual impact, and change its format so that it is better suited for computer processing or human viewing while preventing noise magnification and accomplishing excellent real-time performance. The top goal is to combine data from various modal images (CT/MRI and MR-T1/MR-T2) into a solitary image that, to the greatest degree possible, retains the key characteristics (prominent features) of the source images. The clinical accuracy of medical issues is compromised because innumerable classical fusion methods struggle to conserve all the prominent features of the original images. Furthermore, complex implementation, high computation time, and more memory requirements are key problems of transform domain methods. With the purpose of solving these problems, this research suggests a fusion framework for multimodal medical images that makes use of a multi-scale edge-preserving filter and visual saliency detection. The source images are decomposed using a two-scale edge-preserving filter into base and detail layers. Base layers are combined using the addition fusion rule, while detail layers are fused using weight maps constructed using the maximum symmetric surround saliency detection algorithm. The resultant image constructed by the presumed method has improved objective evaluation metrics than other classical methods, as well as unhindered edge contour, more global contrast, and no ringing effect or artifacts. The methodology offers a dominant and symbiotic arsenal of clinical symptomatic, therapeutic, and biomedical research competencies that have the prospective to considerably strengthen medical practice and biological understanding.

Robust multi-scale weighting-based edge-smoothing filter for single image dehazing

The guided image filter (GIF) and weighted guided image filter (WGIF) are local linear model-based good edge-preserving filters. However, due to fixed regularization parameter, they suffer from halo artifacts (morphological artifacts) in the sharp regions. To overcome this issue, a robust multi-scale weighting-based edge-smoothing filter (RMWEF) for single image dehazing is proposed in this paper. It removes morphological artifacts and over-smoothness strongly and preserves edge information precisely in both flat and sharp regions. The proposed dehaze method has four-steps. First, initial transmission map and atmospheric map are estimated by using a novel dark channel prior (DCP) method. Then, the morphological artifacts of initial transmission map are reduced by using non-local haze line averaging (NL-HLA) method. In the third step, transmission map is refined by using the proposed RMWEF. Finally, the haze free image is restored. Theoretical and experimental analysis proves that the proposed algorithm produce effective dehaze results quicker than the existing methods.

Motion estimation from noisy data with unknown distributions using multi-frame phase-preserving denoising

Phase-based motion estimation method is promising due to its high resolution and wide measurement range. Generally, noise exists in digital images and ruins the constant phase assumption of the same target among different frames. However, traditional image-denoising techniques do not consider the phase and result in inaccurate motion estimation. A novel phase-preserving denoising method considering multiple frames is proposed. In this approach, multi-scale and multi-orientation quadrature filters transform noisy images into complex pyramids to preserve the phase of the content. An instability indicator is designed to evaluate the relative noise intensity at each level of the complex pyramid among multiple frames. The threshold is then estimated to retain the phase of interest and reconstruct a denoised image with unknown noise distributions. Numerical and experimental examples demonstrate that the proposed method effectively preserves the phase of interest and decreases the error of the estimated motion, outperforming other typical denoising methods under various types of noise.

MSCF: Multi-Scale Canny Filter to Recognize Cells in Microscopic Images

Fish fecundity is one of the most relevant parameters for the estimation of the reproductive potential of fish stocks, used to assess the stock status to guarantee sustainable fisheries management. Fecundity is the number of matured eggs that each female fish can spawn each year. The stereological method is the most accurate technique to estimate fecundity using histological images of fish ovaries, in which matured oocytes must be measured and counted. A new segmentation technique, named the multi-scale Canny filter (MSCF), is proposed to recognize the boundaries of cells (oocytes), based on the Canny edge detector. Our results show the superior performance of MSCF on five fish species compared to five other state-of-the-art segmentation methods. It provides the highest F1 score in four out of five fish species, with values between 70% and 80%, and the highest percentage of correctly recognized cells, between 52% and 64%. This type of research aids in the promotion of sustainable fisheries management and conservation efforts, decreases research’s environmental impact and gives important insights into the health of fish populations and marine ecosystems.

A new bearing fault diagnosis method based on improved weighted multi-scale morphological filter and multi-headed self-attention capsule restricted boltzmann network

Intelligent bearing fault diagnosis plays an important role in improving equipment safety and reducing equipment maintenance costs. Noise in the signal can seriously reduce the accuracy of fault diagnosis. To improve the accuracy of fault diagnosis, a novel noise reduction method based on weighted multi-scale morphological filter (WMMF) is proposed. Firstly, Teager energy operator (TEO) is used to amplify the morphological information of the signal. Then, a scale filtering operator using envelope entropy (SFOEE) is proposed to select appropriate scales. At these scales, the noise in the signal can be adequately suppressed. A new weighting method is proposed to integrate the selected scales to construct the WMMF. Finally, multi-headed self-attention capsule restricted boltzmann network (MSCRBN) is proposed to diagnose bearing faults.The performance of the TEO-SFOEE-WMMF-MSCRBN fault diagnosis method is verified on the CWRU dataset. Compared with existing fault diagnosis methods, this approach achieves 100% identification accuracy. The experimental results indicate that the proposed diagnosis method can effectively resist noise and precisely diagnose bearing faults.

Multi-Scale Spatial–Spectral Attention-Based Neural Architecture Search for Hyperspectral Image Classification

Convolutional neural networks (CNNs) are indeed commonly employed for hyperspectral image classification. However, the architecture of cellular neural networks typically requires manual design and fine-tuning, which can be quite laborious. Fortunately, there have been recent advancements in the field of Neural Architecture Search (NAS) that enable the automatic design of networks. These NAS techniques have significantly improved the accuracy of HSI classification, pushing it to new levels. This article proposes a Multi-Scale Spatial–Spectral Attention-based NAS, MS3ANAS) framework for HSI classification to automatically design a neural network structure for HSI classifiers. First, this paper constructs a multi-scale attention mechanism extended search space, which considers multi-scale filters to reduce parameters while maintaining large-scale receptive field and enhanced multi-scale spectral–spatial feature extraction to increase network sensitivity towards hyperspectral information. Then, we combined the slow–fast learning architecture update paradigm to optimize and iteratively update the architecture vector and effectively improve the model’s generalization ability. Finally, we introduced the Lion optimizer to track only momentum and use symbol operations to calculate updates, thereby reducing memory overhead and effectively reducing training time. The proposed NAS method demonstrates impressive classification performance and effectively improves accuracy across three HSI datasets (University of Pavia, Xuzhou, and WHU-Hi-Hanchuan).

M2GF: Multi-Scale and Multi-Directional Gabor Filters for Image Edge Detection

An image edge detection algorithm using multi-directional and multi-scale Gabor filters is proposed in this paper. The main merit of this method is that high edge detection accuracy can be obtained while maintaining noise robustness. The approach proposed in this paper consists of three procedures: firstly, the transformation to the CIE L*a*b* color space, which has a wide shading area and uniform distribution; secondly, under different scales, the edge feature information of the image is extracted from several different directions by Gabor filters, and a new edge strength map is obtained by feature fusion; thirdly, the new fused edge strength map is enhanced with local features, and a noise-resistant image edge detector is obtained under a novel hysteresis threshold calculation. The experiments illustrate that, compared to the methods involved, the designed edge detector outperforms by about 2% to 4%, and also shows competitive performance regarding the ability to handle noise.

A Gaussian multi-scale mixture model-based outlier-robust Kalman filter

In many practical applications, the unknown non-stationary heavy-tailed distributed process and measurement noises (HTPMNs) are prone to occur due to inaccurate noise prior and stochastic outliers. To achieve the state estimation under this situation, a new outlier-robust Kalman filter based on Gaussian multi-scale mixture model (GMSMM-ORKF) is proposed in this paper. First, a new GMSMM is designed to model the one-step prediction and measurement likelihood probability density functions. Then, the hierarchical prior presentation on the mixture probability vectors and scale parameters are built. Furthermore, employing the variational Bayesian (VB) inference, a GMSMM-ORKF is derived. Finally, a classical target tracking model and real navigation experiment are utilized to demonstrate the effectiveness of the proposed filter.

Multiscale Bayesian PCA for robust process modeling of a Fischer–Tropsch bench scale process

Building a good process model is critical for process monitoring technologies which ensure process safety, reliability, and profitability. In this work, we develop a Multiscale Bayesian PCA (MS-BPCA) algorithm and use it to model a real Fischer–Tropsch (FT) pilot plant. We demonstrate the algorithm’s superior and robust performance in dealing with autocorrelation and noise contamination compared to BPCA and Multiscale PCA. The modeling performance of BPCA depends on the accuracy of the likelihood and prior parameters. Therefore, we explore various techniques (multiscale filtering, polynomial approximation, and PCA) to empirically estimate these parameters. Furthermore, we study the limitations of the BPCA modeling algorithm when its assumptions of gaussian data and independent noise are violated. Finally, this work will compare the modeling performance of all techniques (PCA, Bayesian PCA, Multiscale PCA, and Multiscale Bayesian PCA) with real data collected from a state-of-the-art bench-scale Fischer–Tropsch (FT) process.

Image enhancement method based on multi algorithm fusion

This research suggests an image enhancement method based on multi algorithm fusion. The illumination component and reflection component are obtained by morphological closure. On the one hand, multiple inputs of the estimated illumination and the corresponding weights are fused by multi-scale fusion technology. On the other hand, space conversion technology is used. Two adjusted images are obtained based on multi-scale Gaussian filter and adaptive brightness enhancement algorithm. Then details are extracted from these two images through PCA fusion strategy. The reflection component is improved significantly after above operations. Finally, combine the illumination and reflection components. Compared with the classical algorithm, our method can simultaneously achieve brightness improvement, naturalness maintenance and contrast enhancement.

Recurrent Structure Attention Guidance for Depth Super-resolution

Image guidance is an effective strategy for depth super-resolution. Generally, most existing methods employ hand-crafted operators to decompose the high-frequency (HF) and low-frequency (LF) ingredients from low-resolution depth maps and guide the HF ingredients by directly concatenating them with image features. However, the hand-designed operators usually cause inferior HF maps (e.g., distorted or structurally missing) due to the diverse appearance of complex depth maps. Moreover, the direct concatenation often results in weak guidance because not all image features have a positive effect on the HF maps. In this paper, we develop a recurrent structure attention guided (RSAG) framework, consisting of two important parts. First, we introduce a deep contrastive network with multi-scale filters for adaptive frequency-domain separation, which adopts contrastive networks from large filters to small ones to calculate the pixel contrasts for adaptive high-quality HF predictions. Second, instead of the coarse concatenation guidance, we propose a recurrent structure attention block, which iteratively utilizes the latest depth estimation and the image features to jointly select clear patterns and boundaries, aiming at providing refined guidance for accurate depth recovery. In addition, we fuse the features of HF maps to enhance the edge structures in the decomposed LF maps. Extensive experiments show that our approach obtains superior performance compared with state-of-the-art depth super-resolution methods. Our code is available at: https://github.com/Yuanjiayii/DSR-RSAG.

CONTRAST-ENHANCED MICROTOMOGRAPHY FOR VOLUMETRIC ANALYSIS OF MICROSTRUCTURE IN LIGAMENTS AND TENDONS

Tendons and ligaments play an important role to ensure mobility and stability. To correctly understand the characteristics of these fibrous collagenous connective tissues, it is fundamental to highlight their 3D microstructure. In this study a microtomography (microCT) system was used to acquire human hamstring tendons after performing specific preparations to enhance image contrast. Specifically, samples were treated either through chemical dehydration or by 2% of phosphotungstic acid ([Formula: see text]) in water ([Formula: see text]) or in 70% ethanol ([Formula: see text]) solution. Acquired images were elaborated using dedicated techniques based on 3D Hessian multiscale filter so as to highlight the fibrous structure and identify specific geometric features. For any strategy of sample preparation, the proposed approach resulted to be adequate for identifying fascicle features, thus obtaining structures with diameter in the range of 100–600 [Formula: see text] and proper longitudinal alignment. In conclusion, a novel contrast enhancement microCT protocol was designed and preliminarily validated for the microstructural analysis of fibrous tissues.

Computational Modeling of Multiscale Air Filter Media Consisting of Nano- and Microfibers

With regard to air filtration, computational work has been performed to predict filter performance outcomes in different particulate scenarios, yet comparative verification of different modeling methods is rarely studied. In this study, computational simulations with different modeling techniques are demonstrated for filter media with various fiber diameters, thicknesses, and basis weights. For a microscale meltblown (Micro) filter web, a resolved model reconstructed by X-ray micro-computed tomography (Xμ-CT) is generated. The representative volume elements and the number of voxels are determined by examining the simulation accuracy and the computational time. For a multiscale dual-layer web composed of nanofibers (Nano) and microscale spunbond fibers (SB), a parametric model and a porous plane model are generated. The accuracy of the models is verified in terms of the morphological parameters and flow resistance in comparison, with actual test results. The parametric model and porous plane model suitably predict the characteristics of multiscale filter media. Simulated filtration is conducted for particles of different sizes (0.05–1 μm) in an effort to understand the relationships among the filter morphology, face velocity (71 and 142 mm/s), and particle size. This study presents relevant modeling methods, specifically a resolved model, a parametric model, and a porous plane model for virtualizing filter media on various scales. It provides an informative discussion of various modeling parameters for accurate predictions of filtering behaviors with potential applicability to the reverse-engineering of filter products.

An illumination-dependent adaptive fusion method for infrared and visible images

Changes in illumination conditions of the imaging scene cause changes in the relative amount of complementary information between the infrared and visible images. Therefore, the fusion method should adjust to changing illumination conditions to adaptively fuse as much source images’ information as possible. However, most existing fusion methods do not consider the changes in the illumination conditions in the construction process, resulting in the loss of detail information and low contrast in fused images. We propose an illumination-dependent adaptive fusion (IDA Fusion) method to solve this problem. First, to process detail information of different scales and brightness information separately, the multi-scale rolling guidance filter (RGF) is chosen to decompose source images into small-scale detail layers, large-scale detail layers, and base layers. Second, according to the respective characteristics of small-scale and large-scale details under different illumination conditions, two different illumination-dependent rules are designed to combine the small-scale and large-scale detail layers, respectively. These rules adjust their forms and parameters according to the average pixel value and entropy ratio to adaptively transfer the detail information from source to fused images. Moreover, for combining the base layers, a rule based on weighted least squares (WLS) minimization is proposed to keep as much source images’ information as possible while maintaining an appropriate brightness under different illumination conditions. Experimental results validate the effectiveness of the above rules and demonstrate that our method performs better than some state-of-the-art methods, including information retention and contrast improvement.

A Depth-Wise Separable U-Net Architecture with Multiscale Filters to Detect Sinkholes

Numerous variants of the basic deep segmentation model—U-Net—have emerged in recent years, achieving reliable performance across different benchmarks. In this paper, we propose an improved version of U-Net with higher performance and reduced complexity. This improvement was achieved by introducing a sparsely connected depth-wise separable block with multiscale filters, enabling the network to capture features of different scales. The use of depth-wise separable convolution significantly reduces the number of trainable parameters, making the training faster, while reducing the risk of overfitting. We used our developed sinkhole dataset and the available benchmark nuclei dataset to assess the proposed model’s performance. Pixel-wise annotation is laborious and requires a great deal of human expertise; therefore, we propose a fully deep convolutional autoencoder network that utilizes the proposed block to automatically annotate the sinkhole dataset. Our segmentation model outperformed the state-of-the-art methods, including U-Net, Attention U-Net, Depth-Separable U-Net, and Inception U-Net, achieving an average improvement of 1.2% and 1.4%, respectively, on the sinkhole and the nuclei datasets, with 94% and 92% accuracy, as well as a reduced training time. It also achieved 83% and 80% intersection-over-union (IoU) on the two datasets, respectively, which is an 11.8% and 9.3% average improvement over the above-mentioned models.

Self scale estimation of the tracking window merged with adaptive particle filter tracker

Tracking a mobile object is one of the important topics in pattern recognition, but style has some obstacles. A Reliable tracking system must adjust their tracking windows in real time according to appearance changes of the tracked object. Furthermore, it has to deal with many challenges when one or multiple objects need to be tracked, for instance when the target is partially or fully occluded, background clutter, or even some target region is blurred. In this paper, we will present a novel approach for a single object tracking that combines particle filter algorithm and kernel distribution that update its tracking window according to object scale changes, whose name is multi-scale adaptive particle filter tracker. We will demonstrate that the use of particle filter combined with kernel distribution inside the resampling process will provide more accurate object localization within a research area. Furthermore, its average error for target localization was significantly lower than 21.37 pixels as the mean value. We have conducted several experiments on real video sequences and compared acquired results to other existing state of the art trackers to demonstrate the effectiveness of the multi-scale adaptive particle filter tracker.

FFV-MBC: A Novel Fused Finger-Vein Recognition Method Based on Monogenic Binary Coding

<p>To improve pattern representation capabilities and robustness in traditional finger-vein recognition algorithms. In this paper, we propose FFV-MBC, a novel fused finger-vein recognition method based on monogenic binary coding (MBC). First of all, the amplitude, orientation, and phase information of the finger-vein images are filtered by a multi-scale monogenic log-Gabor filter and encoded by the binary coding theory. Three local features, MBC-A, MBC-P, and MBC-O, are achieved from different combinations of local image intensity and variation coding. After obtaining the features, we utilize the block-based Fisher Linear Discriminant method to reduce the dimension. Finally, the similarity components are calculated by the cosine distance and fused for the final finger-vein recognition results. We evaluate our proposed method on two publicly available datasets and one self-built dataset, i.e., Malaysian Polytechnic University (FV-USM), the Group of Machine Learning and Applications of Shandong University (SDUMLA-HMT), and our team, Signal and Information Processing Laboratory (FV-SIPL). On average, the proposed method achieved high recognition accuracy, i.e., 99.30%, and 1.10% equal error rates (EER). Overall, the proposed method performs better than most classical and state-of-the-art finger-vein recognition methods.</p> <p> </p>

Multiscale Partial Symbolic Transfer Entropy for Time-Delay Root Cause Diagnosis in Nonstationary Industrial Processes

Nonstationary characteristics and lack of time delays analysis remain two large obstacles to diagnosing fault root cause in industrial processes. However, ignoring these two issues, many traditional methods tend to introduce some spurious or indirect causal relationships that interfere with root cause judgments. In this article, multiscale partial symbolic transfer entropy (MPSTE) is proposed to handle the above-mentioned problems. On the one hand, a multivariate multiscale filter module of MPSTE is designed to estimate the information transfer on multiple time scales and detect time delays while considering the indirect causality of the multivariate industrial system. On the other hand, to capture the causal information from the nonstationary processes, MPSTE introduces a segmental symbolization procedure and maps time series into finite rank vectors, overcoming the limitation of stationarity. Moreover, two causal criteria for MPSTE are constructed to identify the direction and significance level of causality correspondingly. MPSTE can thus provide sufficient information for causal relationships and exclude spurious causality. Finally, a causal diagram construction strategy with time delay based on MPSTE is established to diagnose the root cause and trace the fault propagation path. The results of experimental verification in a coal mill rig prove the effectiveness of the proposed strategy.