Domain Filtering Research Articles

Phase congruency-based filtering approach combined with a convolutional network for lung CT image analysis

ABSTRACT Computed tomography (CT) imaging was confirmed as one of the important evolving radiological features of disease diagnosis. In this work, we propose an image filtering method for CT images that fuses the spatial and the transform domain. Filters in the transform domain work well in restoring low-contrast details that are usually smoothed by spatial domain filters. The new filter is based on a modified anisotropic diffusion approach combined with the phase congruency (PC) feature. The PC feature is a scale-invariant compared to the classical gradient, where weak edges are usually omitted and undetected with gradient-based feature detectors. This feature is incorporated in the diffusion function to enhance image edges while eliminating noise and texture background. In a further step, a U-net convolutional network is used to segment the lung-infected area. Comparative tests and results show the efficiency of our method, which permits to enhance image features while removing noise perfectly.

Image Denoising Using Nonlocal Regularized Deep Image Prior

Deep neural networks have shown great potential in various low-level vision tasks, leading to several state-of-the-art image denoising techniques. Training a deep neural network in a supervised fashion usually requires the collection of a great number of examples and the consumption of a significant amount of time. However, the collection of training samples is very difficult for some application scenarios, such as the full-sampled data of magnetic resonance imaging and the data of satellite remote sensing imaging. In this paper, we overcome the problem of a lack of training data by using an unsupervised deep-learning-based method. Specifically, we propose a deep-learning-based method based on the deep image prior (DIP) method, which only requires a noisy image as training data, without any clean data. It infers the natural images with random inputs and the corrupted observation with the help of performing correction via a convolutional network. We improve the original DIP method as follows: Firstly, the original optimization objective function is modified by adding nonlocal regularizers, consisting of a spatial filter and a frequency domain filter, to promote the gradient sparsity of the solution. Secondly, we solve the optimization problem with the alternating direction method of multipliers (ADMM) framework, resulting in two separate optimization problems, including a symmetric U-Net training step and a plug-and-play proximal denoising step. As such, the proposed method exploits the powerful denoising ability of both deep neural networks and nonlocal regularizations. Experiments validate the effectiveness of leveraging a combination of DIP and nonlocal regularizers, and demonstrate the superior performance of the proposed method both quantitatively and visually compared with the original DIP method.

Joint Inspection of HD Video and Robot in Substation Based on OCR Technology

The research on the joint inspection technology of HD video and robot in substation is of great significance in the operation and maintenance management of substation. The related research on multidimensional collaborative inspection technology of substation has attracted great attention. The substation monitoring system is designed to realize multidimensional collaborative inspection of HD video and inspection robot. OCR technology is used to improve the abnormal recognition system of inspection robot equipment. According to the collection of video information, DSP comprehensive information processing is carried out, and the detection information is analyzed by frequency domain filtering method through human-computer interaction interface. Meanwhile, the main component features of substation detection components are extracted by pattern recognition method, and the resolution model of similar features of video monitoring images is constructed. It plays an important role in the daily operation and maintenance of substation to construct the reliability visual tracking and identification model and optimize the multidimensional cooperative inspection system model.

Proposal of Single Sideband Modulation Scheme Using Frequency Domain Filtering

With the rapid development of wireless systems, the demand for frequency resources has been increasing in recent years. Therefore, it is necessary to consider the high-quality communication method that efficiently utilizes finite frequency resources. In this paper, Single Sideband 16 Pulse Amplitude Modulation (SSB 16PAM) scheme for the uplink communication is proposed. It transmits data in only Lower Sideband (LSB) without extra Hilbert components. Under Additive White Gaussian Noise (AWGN) channel environment, Bit Error Rate (BER) performance of the proposed scheme is superior by 3 dB in terms of Carrier-to-Noise Ratio (CNR) to 256 Quadrature Amplitude Modulation (256QAM) scheme with the same frequency efficiency and the same Peak-to-Average Power Ratio (PAPR). Our proposed scheme employs the original frequency domain filter on the transmitter side to form an ideal spectrum. The configuration of its process is almost similar to Single Carrier-Frequency Division Multiple Access (SC-FDMA), moreover, half of the input data on the frequency domain is removed. The proposed frequency domain filter produces the SSBmodulated spectrum with a roll-off rate of zero without degrading the BER performance.

Sparse grid-based adaptive noise reduction strategy for particle-in-cell schemes

We propose a sparse grid-based adaptive noise reduction strategy for electrostatic particle-in-cell (PIC) simulations. By projecting the charge density onto sparse grids we reduce the high-frequency particle noise. Thus, we exploit the ability of sparse grids to act as a multidimensional low-pass filter in our approach. Thanks to the truncated combination technique [1–3], we can reduce the larger grid-based error of the standard sparse grid approach for non-aligned and non-smooth functions. The truncated approach also provides a natural framework for minimizing the sum of grid-based and particle-based errors in the charge density. We show that our approach is, in fact, a filtering perspective for the noise reduction obtained with the sparse PIC schemes first introduced in [4]. This enables us to propose a heuristic based on the formal error analysis in [4] for selecting the optimal truncation parameter that minimizes the total error in charge density at each time step. Hence, unlike the physical and Fourier domain filters typically used in PIC codes for noise reduction, our approach automatically adapts to the mesh size, number of particles per cell, smoothness of the density profile and the initial sampling technique. It can also be easily integrated into high performance large-scale PIC code bases, because we only use sparse grids for filtering the charge density. All other operations remain on the regular grid, as in typical PIC codes. We demonstrate the efficiency and performance of our approach with two test cases: the diocotron instability in two dimensions and the three-dimensional electron dynamics in a Penning trap. Our run-time performance studies indicate that our approach can provide significant speedup and memory reduction to PIC simulations for achieving comparable accuracy in the charge density.

Synchronized perturbation elimination and DOA estimation via signal selection mechanism and parallel deep capsule networks in multipath environment

State-of-the-art model-driven Direction-Of-Arrival (DOA) estimation methods for multipath signals face great challenges in practical application because of the dependence on the precise multipath model. In this paper, we introduce a framework, based on deep learning, for synchronizing perturbation auto-elimination with effective DOA estimation in multipath environment. Firstly, a signal selection mechanism is introduced to roughly locate specific signals to spatial subregion via frequency domain filters and compressive sensing-based method. Then, we set the mean of the correlation matrix’s row vectors as the input feature to construct the spatial spectrum by the corresponding single network within the parallel deep capsule networks. The proposed method enhances the generalization capability to untrained scenarios and the adaptability to non-ideal conditions, e.g., lower SNRs, smaller snapshots, unknown reflection coefficients and perturbational steering vectors, which make up for the defects of the previous model-driven methods. Simulations are carried out to demonstrate the superiority of the proposed method.

Inertial Sensor Data to Image Encoding for Human Action Recognition

Convolutional Neural Networks (CNNs) are successful deep learning models in the field of computer vision. To get the maximum advantage of CNN model for Human Action Recognition (HAR) using inertial sensor data, in this paper, we use four types of spatial domain methods for transforming inertial sensor data to activity images, which are then utilized in a novel fusion framework. These four types of activity images are Signal Images (SI), Gramian Angular Field (GAF) Images, Markov Transition Field (MTF) Images and Recurrence Plot (RP) Images. Furthermore, for creating a multimodal fusion framework and to exploit activity images, we made each type of activity images multimodal by convolving with two spatial domain filters: Prewitt filter and High-boost filter. ResNet-18, a CNN model, is used to learn deep features from multi-modalities. Learned features are extracted from the last pooling layer of each ResNet and then fused by canonical correlation based fusion (CCF) for improving the accuracy of human action recognition. These highly informative features are served as input to a multi-class Support Vector Machine (SVM). Experimental results on three publicly available inertial datasets show the superiority of the proposed method over the current state-of-the-art.

LineSmooth: An Analytical Framework for Evaluating the Effectiveness of Smoothing Techniques on Line Charts.

We present a comprehensive framework for evaluating line chart smoothing methods under a variety of visual analytics tasks. Line charts are commonly used to visualize a series of data samples. When the number of samples is large, or the data are noisy, smoothing can be applied to make the signal more apparent. However, there are a wide variety of smoothing techniques available, and the effectiveness of each depends upon both nature of the data and the visual analytics task at hand. To date, the visualization community lacks a summary work for analyzing and classifying the various smoothing methods available. In this paper, we establish a framework, based on 8 measures of the line smoothing effectiveness tied to 8 low-level visual analytics tasks. We then analyze 12 methods coming from 4 commonly used classes of line chart smoothing-rank filters, convolutional filters, frequency domain filters, and subsampling. The results show that while no method is ideal for all situations, certain methods, such as Gaussian filters and TOPOLOGY-based subsampling, perform well in general. Other methods, such as low-pass CUTOFF filters and Douglas-peucker subsampling, perform well for specific visual analytics tasks. Almost as importantly, our framework demonstrates that several methods, including the commonly used UNIFORM subsampling, produce low-quality results, and should, therefore, be avoided, if possible.

Bilateral Filter for the Optimization of Composite Structures

In the present study, we propose to integrate the bilateral filter into the Shepard-interpolation-based method for the optimization of composite structures. The bilateral filter is used to avoid defects in the structure that may arise due to the gap/overlap of adjacent fiber tows or excessive curvature of fiber tows. According to the bilateral filter, sensitivities at design points in the filter area are smoothed by both domain filtering and range filtering. Then, the filtered sensitivities are used to update the design variables. Through several numerical examples, the effectiveness of the method was verified.

Partial Color Photo Processing Method for Components Based on Image Enhancement Technology

Computer vision is currently playing an increasingly important role in automatically identifying the character of the image processing technology as research hotbed in the field of smart computing, OCR, face recognition, fingerprinting, biometric recognition, and so forth. Content‐based image recovery, video recovery, multimedia collection, watermarking, games, film stunts, virtual reality, e‐commerce, and other apps are available all round. The color pictures of parts taken by industrial cameras depend on computer performance and the intricate environment, and in particular, on the whole resolution image display, a lot of CPU resources are needed. Some details cannot be shown completely at the same time. If the image is not sufficiently clearly visible, methods for image processing like improvement, noise reduction, and interpolation must be used to improve color photo clarity. This article, based on the OpenCV platform, uses frequency domain filters, median filters, Fourier transform, and other image improvement technologies to remove image noise in order to enhance the quality of local photos from industrial cameras’ components. Finally, clear and available image information is obtained in different experimental methods, which check the application of image enhancement technology to image rebuilding. Finally, the performance of the proposed method in terms of CPBD value, definition Q value, and operation time is compared, which shows that the proposed method has obvious advantages in the above performance.

An Effective Low-Contrast SF₆ Gas Leakage Detection Method for Infrared Imaging

Infrared noncontact detection has the advantages of having no power cut-off and convenient operation. However, when the SF6 (sulfur hexafluoride) gas leakage is small, the existing detection algorithms have low accuracy and are easily disturbed by noise. To solve this problem, this article proposes an online detection method for SF6 gas leakage based on a Gaussian mixture model. First, as a preprocessing stage, the improved dynamic interframe time domain filter is used to suppress the random noise in the image. Then, contrast limited adaptive histogram equalization (CLAHE) is used to enhance the dark details of the infrared image to improve the local contrast. Finally, the improved Gaussian mixture background model is used to adaptively segment the SF6 gas leakage and mark the leakage area. The results show that under the experimental conditions of 0.06 mL/min indoor and 5-m distance, the algorithm can overcome the high noise and complex background disturbances of infrared imaging. It can effectively detect and locate the SF6 leakage area and gets a higher F1-score compared with other conventional algorithm under the experimental conditions. In addition, when the resolution of the infrared image is $320 \times 240$ , the running efficiency of the algorithm meets the real-time requirements.

Calibration of hydrophones using a frequency domain filter processing method: theory and experiment

To eliminate the interferences coming from the boundaries of a water tank and extend its low frequency calibration limit, a frequency domain filter processing (FDFP) method is proposed for hydrophone calibration measurements. The frequency domain filter function is designed according to the a priori spatial position of projector and hydrophone pair in a water tank. A simultaneous comparison method is used for calibration, and a wideband pink noise pulse signal is transmitted. After application of the FDFP method, the transfer impedances of projector and hydrophone pairs in a quasi-free-field are obtained, and interferences reflected from the boundaries of the water tank are effectively eliminated. To verify the FDFP method, an experimental system is established for calibration of the free-field sensitivity of a TC 4033 hydrophone. For a non-anechoic water tank of 1.5 m × 0.88 m × 0.90 m, the lowest calibration frequency is extended from 20 kHz to 3.15 kHz, and the calibration result is consistent with the free-field reciprocity calibration result obtained in an anechoic water tank of 50 m × 15 m × 10 m in the frequency range 3.15 kHz to 31.5 kHz.

Influence of Neighborhood Size and Cross-Correlation Peak-Fitting Method on Location Accuracy

A known technique to obtain subpixel resolution by using object tracking through cross-correlation consists of interpolating the obtained correlation function and then refining peak location. Although the technique provides accurate results, peak location is usually biased toward the closest integer coordinate. This effect is known as the peak-locking error and it strongly limits this calculation technique’s experimental accuracy. This error may differ depending on the scene and algorithm used to fit and interpolate the correlation peak, but in general, it may be attributed to a sampling problem and the presence of aliasing. Many studies in the literature analyze this effect in the Fourier domain. Here, we propose an alternative analysis on the spatial domain. According to our interpretation, the peak-locking error may be produced by a non-symmetrical sample distribution, thus provoking a bias in the result. According to this, the peak interpolant function, the size of the local domain and low-pass filters play a relevant role in diminishing the error. Our study explores these effects on different samples taken from the DIC Challenge database, and the results show that, in general, peak fitting with a Gaussian function on a relatively large domain provides the most accurate results.

Chirp multiplexing and filtering in the offset linear canonical transform domain

This paper shows how offset linear canonical transform (OLCT) can be handled for multiplexing chirp signals in time–frequency domain in the presence of additive white Gaussian noise (AWGN). A set of original test signals have been processed in this framework for OLCT and offset fractional Fourier transform (OFRFT) modalities. After recovering, the resulting signals obtained from these two modalities have been evaluated by measuring the mean square error (MSE) to input signals for different signal to noise ratio (SNR) values and found MSE is minimum in case of LCT multiplexing and filtering domain as compared to the corresponding fractional domain.

Random noise reduction using SVD in the frequency domain

The frequency spectrum of irregular interference noise has broad bandwidth and poor coherence.But in the same prospecting area, the dominant frequency and bandwidth of effective signals are nearly the same (especially for post-stack section); that is to say, the frequency spectra of effective signals in seismic traces show a high degree of trace-to-trace correlation. Based on this conclusion, we present a novel denoising technique, which works by SVD filtering in the frequency domain. First, the input seismic data are transformed to the frequency domain via the Fourier transform. Then, the frequency spectra are decomposed into eigenimages by means of SVD. We perform the eigenimage filtering of the frequency spectra by selecting singular values to be used in the reconstruction, suppressing the random noise. Compared with the traditional band-pass filtering, the presented method is capable of attenuating the interference noise components within the range of frequency pass band and protects effective signals in high frequency. Tests on both synthetic and field seismic data show that our method can remove random noise and does no damage to effective signal. By comparison with the median filtering and the curvelet domain filtering, we concluded that the presented denoising method performs better in removing background noise and protecting reflection events.

Novel Deep Level Image State Ensemble Enhancement Method for M87 Imaging

Standard spatial domain filters fail to adequately denoise and enhance the contrast of an image. These filters have drawbacks like oversmoothing, diminished texture, and lack of generative capabilities. This paper proposes a new method of image reconstruction, Image State Ensemble Enhancement (ISEE), based on our previous work, Image State Ensemble Decomposition (ISED). Deep level ISEE and ISED have been developed to produce a class of filters that can address these issues. Full-reference and no-reference quality metrics are used to assess the image, and the full reference metrics showed a marked improvement, while the no-reference metrics were often better than the test image. The test image was taken from the Spitzer Space Telescope (SST), and ISEE reconstruction yielded improved structural detail over that of ISED and the original test image. Glare and noise were reduced in a narrow bandwidth, which led to the discovery of a vortex-shaped structure and an outburst in M87′s dusty infrared core. The vortex is located over M87′s visible core and black hole. This is verified with an SST and Hubble Space Telescope (HST) overlay, ISEE processed image. A counter-jet channel was also discovered, and it appears to be the path of the unobservable superluminal counter-jet.

Computationally Efficient Wildfire Detection Method Using a Deep Convolutional Network Pruned via Fourier Analysis.

In this paper, we propose a deep convolutional neural network for camera based wildfire detection. We train the neural network via transfer learning and use window based analysis strategy to increase the fire detection rate. To achieve computational efficiency, we calculate frequency response of the kernels in convolutional and dense layers and eliminate those filters with low energy impulse response. Moreover, to reduce the storage for edge devices, we compare the convolutional kernels in Fourier domain and discard similar filters using the cosine similarity measure in the frequency domain. We test the performance of the neural network with a variety of wildfire video clips and the pruned system performs as good as the regular network in daytime wild fire detection, and it also works well on some night wild fire video clips.

An improved common spatial pattern combined with channel-selection strategy for electroencephalography-based emotion recognition

Emotional human-computer interaction (HCI) has become an important research area in the fields of artificial intelligence and cognitive science, owing to the requirement for active emotion perception. To enhance the performance of electroencephalography (EEG)-based emotional HCI, this paper proposes an improved common spatial pattern combined with a channel-selection strategy (ICSPCS) for EEG-based emotion recognition. Specifically, we first use a common spatial pattern algorithm to design a spatial domain filter according to three different emotions (positive, neutral, and negative). The traditional joint approximation diagonalization method using the criterion of the “highest score eigenvalue” may be unable to solve multiple classifications of emotion representation. Therefore, we design three different eigenvalue selection methods in terms of the positions of the eigenvalues with the highest scores. Finally, to make the installation easier and reduce the computational load, we also develop a channel-selection strategy based on the weight values that individually reflect the degrees of influence of all the channels on emotion recognition. Experiments are conducted on a self-collected dataset and on the MAHNOB-HCI dataset. The average recognition accuracies for the three emotion tasks are found to be 85.85% and 94.13% on the self-collected and MAHNOB-HCI datasets, respectively, thus proving the effectiveness of the proposed ICSPCS method for emotion recognition.

Despeckling of polarimetric SAR images using optimized bandelet transform and GCV thresholding

Polarimetric synthetic aperture radar (PolSAR) imaging opened up a new era in remote sensing by exploiting the advantages of electromagnetic signals over optical signals. Much hype and hope exists over optical imaging due to its all weather imaging capability and capturing finer target details. However as in SAR imaging, PolSAR is also affected by speckle noise which is generally treated as multiplicative in nature. In order to extract target features, it is inevitable to filter out the speckle noise as part of preprocessing step in any image processing application. Transform domain filtering is desirable over spatial domain techniques due to the energy compaction property and easiness of filtering using thresholding. A novel method of despeckling using bandelet transform is proposed here as bandelets retain the geometrical features of the target effectively in comparison with other transforms post processing. Since the computational complexity involved with bandelet transform is relatively high compared to the state of the art techniques, a new approach using optimized bandelet transform based on directional variance coupled with optimal GCV threshold is proposed here. Performance comparison using both air borne and space borne real PolSAR images show that the proposed technique is much more effective and computationally efficient.

Novel Image State Ensemble Decomposition Method for M87 Imaging

This paper proposes a new method of image decomposition with a filtering capability. The image state ensemble decomposition (ISED) method has generative capabilities that work by removing a discrete ensemble of quanta from an image to provide a range of filters and images for a single red, green, and blue (RGB) input image. This method provides an image enhancement because ISED is a spatial domain filter that transforms or eliminates image regions that may have detrimental effects, such as noise, glare, and image artifacts, and it also improves the aesthetics of the image. ISED was used to generate 126 images from two tagged image file (TIF) images of M87 taken by the Spitzer Space Telescope. Analysis of the images used various full and no-reference quality metrics as well as histograms and color clouds. In most instances, the no-reference quality metrics of the generated images were shown to be superior to those of the two original images. Select ISED images yielded previously unknown galactic structures, reduced glare, and enhanced contrast, with good overall performance.