Anisotropic Filter Research Articles

Corner Detection Using Multi-directional Structure Tensor with Multiple Scales

Corners are important features for image analysis and computer vision tasks. Local structure tensors with multiple scales are widely used in intensity-based corner detectors. In this paper, the properties of intensity variations of a step edge, L-type corner, Y- or T-type corner, X-type corner, and star-type corner are investigated. The properties that we obtained indicate that the image intensity variations of a corner are not always large in all directions. The properties also demonstrate that existing structure tensor-based corner detection methods cannot depict the differences of intensity variations well between edges and corners which result in wrong corner detections. We present a new technique to extract the intensity variations from input images using anisotropic Gaussian directional derivative filters with multiple scales. We prove that the new extraction technique on image intensity variation has the ability to accurately depict the characteristics of edges and corners in the continuous domain. Furthermore, the properties of the intensity variations of step edges and corners enable us to derive a new multi-directional structure tensor with multiple scales, which has the ability to depict the intensity variation differences well between edges and corners in the discrete domain. The eigenvalues of the multi-directional structure tensor with multiple scales are used to develop a new corner detection method. Finally, the criteria on average repeatability (under affine image transformation, JPEG compression, and noise degradation), region repeatability based on the Oxford dataset, repeatability metric based on the DTU dataset, detection accuracy, and localization accuracy are used to evaluate the proposed detector against ten state-of-the-art methods. The experimental results show that our proposed detector outperforms all the other tested detectors.

Anisotropic Guided Filtering.

The guided filter and its subsequent derivatives have been widely employed in many image processing and computer vision applications primarily brought about by their low complexity and good edge-preservation properties. Despite this success, the different variants of the guided filter are unable to handle more aggressive filtering strengths leading to the manifestation of "detail halos". At the same time, these existing filters perform poorly when the input and guide images have structural inconsistencies. In this paper, we demonstrate that these limitations are due to the guided filter operating as a variable-strength locally-isotropic filter that, in effect, acts as a weak anisotropic filter on the image. Our analysis shows that this behaviour stems from the use of unweighted averaging in the final steps of guided filter variants including the adaptive guided filter (AGF), weighted guided image filter (WGIF), and gradient-domain guided image filter (GGIF). We propose a novel filter, the Anisotropic Guided Filter (AnisGF), that utilises weighted averaging to achieve maximum diffusion while preserving strong edges in the image. The proposed weights are optimised based on the local neighbourhood variances to achieve strong anisotropic filtering while preserving the low computational cost of the original guided filter. Synthetic tests show that the proposed method addresses the presence of detail halos and the handling of inconsistent structures found in previous variants of the guided filter. Furthermore, experiments in scale-aware filtering, detail enhancement, texture removal, and chroma upsampling demonstrate the improvements brought about by the technique.

Assessment of Solution Algorithms for LES of Turbulent Flows Using OpenFOAM

We validate and test two algorithms for the time integration of the Boussinesq form of the Navier—Stokes equations within the Large Eddy Simulation (LES) methodology for turbulent flows. The algorithms are implemented in the OpenFOAM framework. From one side, we have implemented an energy-conserving incremental-pressure Runge–Kutta (RK4) projection method for the solution of the Navier–Stokes equations together with a dynamic Lagrangian mixed model for momentum and scalar subgrid-scale (SGS) fluxes; from the other side we revisit the PISO algorithm present in OpenFOAM (pisoFoam) in conjunction with the dynamic eddy-viscosity model for SGS momentum fluxes and a Reynolds Analogy for the scalar SGS fluxes, and used for the study of turbulent channel flows and buoyancy-driven flows. In both cases the validity of the anisotropic filter function, suited for non-homogeneous hexahedral meshes, has been studied and proven to be useful for industrial LES. Preliminary tests on energy-conservation properties of the algorithms studied (without the inclusion of the subgrid-scale models) show the superiority of RK4 over pisoFoam, which exhibits dissipative features. We carried out additional tests for wall-bounded channel flow and for Rayleigh–Bènard convection in the turbulent regime, by running LES using both algorithms. Results show the RK4 algorithm together with the dynamic Lagrangian mixed model gives better results in the cases analyzed for both first- and second-order statistics. On the other hand, the dissipative features of pisoFoam detected in the previous tests reflect in a less accurate evaluation of the statistics of the turbulent field, although the presence of the subgrid-scale model improves the quality of the results compared to a correspondent coarse direct numerical simulation. In case of Rayleigh–Bénard convection, the results of pisoFoam improve with increasing values of Rayleigh number, and this may be attributed to the Reynolds Analogy used for the subgrid-scale temperature fluxes. Finally, we point out that the present analysis holds for hexahedral meshes. More research is need for extension of the methods proposed to general unstructured grids.

Anisotropic Joint Trilateral Rolling Filter for Image Smoothing

Filtering an image by eliminating noise and irrelevant details while preserving prominent structure edges is an important pre-processing task in many fields, such as image processing and computer vision. In this study, a novel approach called anisotropic joint trilateral rolling filter (AJTRF) is proposed for the smoothing of an image while preserving important structure edges. AJTRF works in an iterative manner and extends the joint bilateral filter to a joint trilateral filter by employing additional weights in eigen space. During filtering, the range weights are recalculated in each iteration, while the spatial Gaussian weights are computed in anisotropic orientations instead of isotropic orientations. Furthermore, the inner products between anisotropic orientations (expressed as eigenvectors) are used as weights to measure orientation similarities of pixels. Structure tensors at each pixel are calculated to determine the structure orientations, which are used as the anisotropic orientations. A range of experiments are conducted. The results demonstrate that the proposed AJTRF has stronger smoothing and structure preserving ability than established methods.

Direct anisotropic filter method of generating synthetic turbulence applied to turbulence-airfoil interaction noise prediction

Synthetic turbulence methods are widely used in computational aeroacoustics to study broadband noise physics. In this work, a direct anisotropic filter (DAF) method is proposed to produce desired velocity spectra of three-dimensional anisotropic turbulence, based on a direct form of anisotropic filter that can be implemented directly to obtain non-Gaussian spectra. It contains fewer model constraints than the existing methods, getting rid of the inconvenience induced by complex hypergeometric functions and Gaussian superposition with Mach number dependent model parameters. In addition, compared with the numerical results given by existing methods, e.g., the advanced digital filter method, this technique offers improvement in accuracy and cost. In this study, the DAF method is combined with the linearised Euler equations in computational aeroacoustics to study the turbulence-airfoil interaction noise, with the results showing good agreement with benchmarking analytical solutions and experimental measurement. The numerical study also investigates the effect of anisotropy on leading edge noise for thick airfoils, revealing that the anisotropic turbulence stretched in the axial direction will suppress the thickness-induced noise reduction. This study also highlights the contribution of streamwise energy spectra to noise generation when considering the effects of anisotropy.

Structural adaptive anisotropic recursive filter for blind medical image deconvolution

Performance of radiographic diagnosis and therapeutic intervention heavily depends on the quality of acquired images. Over decades, a range of pre-processing for image enhancement has been explored. Among the most recent proposals is iterative blinded image deconvolution, which aims to identify the inheritant point spread function, degrading images during acquisition. Thus far, the technique has been known for its poor convergence and stability and was recently superseded by non-negativity and support constraints recursive image filtering. However, the latter requires a priori on intrinsic properties of imaging sensor, e.g., distribution, noise floor and field of view. Most importantly, since homogeneity assumption was implied by deconvolution, recovered degrading function was global, disregarding fidelity of underlying objects. This paper proposes a modified recursive filtering with similar non-negativity constraints, but also taking into account local anisotropic structure of content. The experiment reported herein demonstrates its superior convergence property, while also preserving crucial image feature.

Structural adaptive anisotropic recursive filter for blind medical image deconvolution

Performance of radiographic diagnosis and therapeutic intervention heavily depends on the quality of acquired images. Over decades, a range of pre-processing for image enhancement has been explored. Among the most recent proposals is iterative blinded image deconvolution, which aims to identify the inheritant point spread function, degrading images during acquisition. Thus far, the technique has been known for its poor convergence and stability and was recently superseded by non-negativity and support constraints recursive image filtering. However, the latter requires a priori on intrinsic properties of imaging sensor, e.g., distribution, noise floor and field of view. Most importantly, since homogeneity assumption was implied by deconvolution, recovered degrading function was global, disregarding fidelity of underlying objects. This paper proposes a modified recursive filtering with similar non-negativity constraints, but also taking into account local anisotropic structure of content. The experiment reported herein demonstrates its superior convergence property, while also preserving crucial image feature.

A robust anisotropic diffusion filter with low arithmetic complexity for images

Image smoothing with edge preservation in the presence of outliers is a challenge in image processing. Anisotropic diffusion smoothing is a well-established paradigm in digital image smoothing with edge preservation. Anisotropic diffusion smoothing filters are not robust to impulse noise. They employ two or more stages for robustness in the presence of outliers, and, therefore, they demand significantly increased arithmetic operations for robustness; consequently, they are not power efficient. This paper introduces a low arithmetic complexity image smoothing model and proposes an intrinsically robust and power efficient algorithm for anisotropic diffusion smoothing of images. The algorithm outperforms the foundational robust smoothing algorithms in terms of the standard performance metrics and visual quality.

Novel Technology for Lung Tumor Detection Using Nanoimage

Nowadays, nanotechnology is gaining more advantages and widely used in many real-life applications including minute tumor detection and effective diagnosis. Nanoscale imaging technique significantly increases the precision, accuracy of tumor detection, and classification of tumor into benign and malignant. Computerized tomography (CT) scan is important for lung cancer diagnosis and research, because it gives accurate segmentation results in lung tumor. In the proposed NBDS method, nanotechnology based detection scheme is used to detect the lung tumors in nanoscale range. Pulmonary nodules are symptoms of lung cancer. The shape and size of these pulmonary nodules are used to diagnose lung cancer in CT images. In the early stages, nodules are very small, and radiologist has to refer to many CT images to diagnose the disease, causing operator mistakes. Image processing algorithms are used as an aid to detect and localize nodules. Here, the input nanoimage is enhanced by using the technique of unsharp masking with the anisotropic filter. By using toboggan algorithm, lung cancer images are segmented. Nano measuring tool for graphical user interface is developed in MATLAB software to detect the lung tumor area or lung lesion in the body (in nanometers). Image classification and feature extraction are done by K-nearest neighbor (KNN) and support vector machine (SVM) with Bag of Visual Words (BoVW) classifiers. The overall accuracy of 97% is obtained using GLCM and FOS features in MATLAB software. The convolutional neural network (CNN) classifier gives the maximum accuracy of 98.19% in MATLAB 2017a software, and hence, this classifier does not need the feature extraction step. Furthermore, the average time consumption for one lesion segmentation was under 4 s using our proposed method. The self-assembled biocompatible nano technique is created to detect the tumor area (in nanometers), and it automatically evaluates the disease. This novel nanotechnology-based tumor area detection schemes achieve robust, efficient, and accurate lung lesion segmentation in nano CT images.

Selective Light Field Refocusing for Camera Arrays Using Bokeh Rendering and Superresolution

Camera arrays provide spatial and angular information within a single snapshot. With refocusing methods, focal planes can be altered after exposure. In this letter, we propose a light field refocusing method to improve the imaging quality of camera arrays. In our method, the disparity is first estimated. Then, the unfocused region (bokeh) is rendered by using a depth-based anisotropic filter. Finally, the refocused image is produced by a reconstruction-based superresolution approach where the bokeh image is used as a regularization term. Our method can selectively refocus images with focused region being superresolved and bokeh being aesthetically rendered. Our method also enables postadjustment of depth of field. We conduct experiments on both public and self-developed datasets. Our method achieves superior visual performance with acceptable computational cost as compared to other state-of-the-art methods. Code is available at https://github.com/YingqianWang/Selective-LF-Refocusing.

Damage detection in carbon fiber–reinforced polymer composite via electrical resistance tomography with Gaussian anisotropic regularization

Electrical resistance tomography is a method for sensing the spatial distribution of electrical conductivity. Therefore, this type of tomography is suitable for sensing damages, which affect electrical conductivity. The utilization of resistance tomography for the structural health monitoring of carbon fiber–reinforced polymer composites is questionable owing to its low spatial resolution and the strong anisotropy of carbon fiber–reinforced polymer composites. This article deals with the employment of resistance tomography with regularization based on a Gaussian anisotropic smoothing filter for the detection of cuts. The advantages of the filter are shown through the image reconstruction of rectangular composite specimens with three different laminate stacking sequences. The cuts are implemented by a milled groove. Visual comparison of the images shows a substantial improvement in the shape reconstruction ability. In addition to visual comparison, the image reconstructions are assessed in terms of the reconstruction error and cross-correlation.

Feature-convinced mesh denoising

Conventional mesh denoising methods suffer from confusing sharp edges with smooth ones, especially for meshes with large noise. Although many recent methods try to make pre-processing, they still lack robustness near feature edges. To address this issue, we propose a new, feature-convinced mesh denoising method. Instead of identifying feature vertices or feature faces, we recognize feature edges in an evolutional manner. To reduce the impact of large noise, we refine a copy of the noisy mesh with Laplacian smoothing and then denoise it with anisotropic filtering. Feature edges and candidate feature lines are formed on the refined mesh. A feature saliency measure is defined to reset feature. The reliable feature information is mapped to the original noisy mesh and feature-aware neighborhoods are constructed to estimate guided normals. Both tangential and normal noise are attenuated simultaneously. Noise level is estimated beforehand according to two statistics and most parameters are set automatically.

Cancelable palmprint templates based on random measurement and noise data for security and privacy-preserving authentication

In order to achieve biometric security and enhance privacy-preserving, a novel palmprint template protection scheme based on random comparison and noise data is proposed. Firstly, Anisotropic Filter (AF) is employed to capture the orientation information of the palmprint. Then, the orientation feature of palmprint is measured by a chaotic matrix to generate secure and cancelable palmprint template. The pseudo-randomness and non-ergodicity of the chaotic matrix can guarantee the security of template. Finally, in order to enhance the privacy protection of the template, the noise data with independent and identically distributed is added, as the final cancelable palmprint template. Theoretical analysis shows that a proper amount of noise has little effect on the recognition accuracy while the privacy is enhanced. During the matching stage, the recognition accuracy can be improved by fusing matching scores at the score-layer or the decision-layer. The theoretical effect of adding noise on the performance is also analyzed. The matching scores of the experimental results are consistent with the theoretical values, which means that we can reasonably adjusted the proportion of noise data through calculations and protect palmprint privacy on the basis of ensuring the recognition accuracy. Furthermore, our methods can still achieve very high security in the worst case of secret key theft.

Designing an anisotropic noise filter for measuring critical dimension and line edge roughness from scanning electron microscope images

The scanning electron microscope (SEM) is often employed in inspecting patterns transferred through a lithographic process. A typical inspection is to measure the critical dimension (CD) and line edge roughness (LER) of each feature in a transferred pattern. Such inspection may be done by utilizing image processing techniques to detect the boundaries of a feature. Since SEM images tend to include a substantial level of noise, a proper reduction of noise is essential before the subsequent process of edge detection. In a previous study, a method of designing an isotropic Gaussian filter adaptive to the noise level was developed. However, its performance for relatively small features was not so good as for large features, especially in the case of LER. The main objective of this study is to improve the design method such that the accuracy of the measured CD and LER is not deteriorated substantially as the feature size decreases. The new design method allows a Gaussian filter to be anisotropic for the better adaptability to the signal and noise, both of which show a substantial level of directional correlation. The cutoff frequency for the direction normal to features is determined to include most of the signal components, and the cutoff frequency in the other direction is set to balance the signal and noise components to be included. This procedure enables a systematic and easy design of the filter. Also, the method of estimating the noise has been modified for higher accuracy. The performance of the new design method has been thoroughly analyzed using the reference images for which feature boundaries are known. It has been shown that the anisotropic filter designed by the proposed method performs better than the isotropic filter from the viewpoint of CD, LER, and power spectral density accuracy.

Texture pattern classification based on probability density function estimation of the image spatial structure feature with symmetrical weibull distribution model

Automatic texture pattern classification (ATPC) has long been an essential issue in computer vision. However, ATPC is still a challenging task since texture is a subjective conception, which is difficult to be expressed concisely by the existing computational models. The visual appearance of the imaged texture pattern (TP) visually depends on the random organization of local homogeneous fragments (LHFs) in it. Hence, it is essential to investigate the latent statistical distribution (LSD) behavior of LHFs for the texture image spatial structural feature (TISSF) characterization and expression to achieve a good performance of ATPC. We presented a probability density function estimation (PDFE)-based ATPC scheme, termed PDFE-ATPC. We demonstrated the Weibull distribution (WD) behavior of LHFs by the sequential fragmentation theory to explain the LSD of the TISSFs. To obtain the multiscale and multi-orientation detail expression of the TISSFs, we introduced an oriented Gaussian derivative filter (OGDF)-based TISSF characterization method, including the steerable isotropic Gaussian derivative filters (SIGDFs) and the oriented anisotropic Gaussian Derivative filters (OAGDFs). Successively, the LSDs of the filter responses were characterized omnidirectionally by a symmetrical WD model (SWDM) and the SWDM-based TISSF parameters, demonstrated to be directly related to the human vision perception (HVP) system, were extracted and applied to the ATPC with a spline regression-based classifier. Effectiveness of the proposed PDFE-ATPC method was verified by extensive experiments on three different texture image databases and compared with four commonly-used statistics-based texture classification methods.

"Fuzzy Qualitative Reasoning Model for Astrocytoma Brain Tumor Grade Diagnosis"

Background: Magnetic Resonance Imaging (MRI) is the most prominently used image acquisition method for brain tumor diagnosis, treatment and research. Objective: In this paper, a fuzzy qualitative reasoning model for diagnosing the grade of Astrocytoma brain tumor using various subtypes of MR images (T1, T1c+, T2, Flair) is explained with its implementation details. Methods: The fuzzy model is implemented in 5 stages namely preprocessing, segmentation, feature extraction, feature selection and building a Fuzzy Inference System (FIS) for diagnosis. In preprocessing, anisotropic filtering is used to remove noise and artifacts whereas the edge information and smoothness are retained. Then the tumor region is segmented by applying active contour method. From the segmented tumor region, textural and shape features are extracted and stored along with the clinical parameters like age, gender and mass effect of the patient for feature selection. The features are analyzed in different dimensions like image, patient, patient with subtype, to determine the sensitive feature subset and its range that discriminates the grade of the tumor. Based on this outcome a Mamdani based fuzzy qualitative reasoning model is built with optimal rule set for tumor grade diagnosis. Findings: The constructed fuzzy model is validated using real data set of MR images and clinical report of patients. The grade of tumor identified is same as that specified in the patient\'s report and hence the model provides better accuracy. Novelty: The novelty of this research work are: subtypes of MR images with analysis in different dimensions, identification of optimal rule set (minimum number of rules without ambiguity), recognition of irregular shape tumor, suitable model for any knowledge based diagnosis. Keywords: Active Contour Method, Anisotropic Filtering, Astrocytoma Brain Tumor, Fuzzy Qualitative Reasoning Model, Magnetic Resonance Images, Optimal Rule Set, Textural and Shape Features

Research on image de‐disturbing algorithm based on dark channel prior and anisotropic Gaussian filtering

SummaryIn order to solve the problem of serious degradation of images collected outdoors in dense fog weather, a defogging algorithm for dense fog images was proposed. The fog‐day imaging physical model was simplified; the concept of fog concentration factor was proposed. The single image de‐hazing algorithm based on dark channel priors solves the problem of estimating the transmittance of fog and sky scenes. The image recovered by the algorithm is clear and natural. The algorithm has high computational complexity. It takes a long time to meet real‐time requirements. The principle of anisotropic Gaussian filtering is introduced. It combines with the dark channel. The value of the fog concentration coefficient is obtained by estimating the visibility value of a single fog image. It is then combined with an anisotropic Gaussian filter of the image to estimate the atmospheric light value. Defogging of the fog image is performed by using the repair function. It is necessary to perform effective de‐hazing processing on smog images. Experiments show that the improved algorithm can greatly reduce the complexity of the algorithm while ensuring the defogging effect of the original algorithm.

Integration of Biomass Gasification and Hot Gas Cleaning Processes

In the near future, most of the world population will live in the developing countries, and it is plausible that biomass will be one of the main renewable energy sources of the future. Fluidized bed reactors should be the best solution to transform biomasses having different physical properties in energy and chemical vectors through gasification, due to its possibility to operate continuously, at high temperatures, and to utilize catalysts into the reactor and downstream of it, in order to reduce tar, NH3 etc. in the product gas.Catalytic filter candles are an innovative solution for hot gas cleaning & conditioning. It has been demonstrated that catalytic ceramic filters inserted in the freeboard of a fluidized bed gasifier, allow the complete removal of particulate by means of their anisotropic porous filtering structure, and furthermore act as catalyst to remove tar and ammonia, thanks to the Ni based catalyst contained into the filter itself (Rapagna, Gallucci and Foscolo, 2017).To by-pass the preparation of the catalytic ceramic filters and to render the overall process more feasible in practice, ceramic filters filled with commercial steam reforming catalyst pellets have been located in the freeboard of a 0.1 m ID fluidized bed reactor, where olivine particles act as bed inventory.Experimental tests were carried out to check the values of the gas pressure drop through the plain filter and through the filter containing catalyst pellets, at different temperatures and gas filtration velocities. It has been found that at the highest temperature and filtration velocity (T 800°C and v 140 m/h) the pressure drops are around 48 mbar, while at the operating conditions (T 800°C and v 100 m/h) the total pressure drops are only 35 mbar. Furthermore it was observed that there is an almost negligible difference between the pressure drops in the case of empty and filled candle; it was thus deduced that the catalyst pellets do not cause relevant additional losses compared to the empty filter. The trends of pressure drops were fitted with the empirical Darcy-Forchheimer relation and with the Ergun equation, and it was found that the major contribution is given by the viscous term of the relation. Furthermore an additional contribution of concentrated pressure drops was identified and ascribed to the outer Al2O3 membrane of the candle characterized by very fine pores.

A survey of denoising techniques for multi-parametric prostate MRI

Denoising is one of active area of research in the image-processing domain since last decade. Internal and external conditions of acquisition device are the main source of noise in an image during the procurement process, which is often impossible to avoid in practical situations. Since many different image denoising algorithms have been recommended till date, but the issue of noise elimination remains an undefended challenge. The main objective of this paper is to study and analyze the behavior of different denoising filters for multi-parametric (mp) prostate MRI so that the appropriate filter can be selected unanimously. This study evaluates the performance of fifteen denoising filters (Anisotropic, Median, Wiener, Gaussian, Mean, Wavelet, Contourlet, Bilateral, Curvelet, WHMT, NLM, GFOE, LMMSE, CURE-LET and ARF) w.r.t mp-prostate MRI i.e. T2w, DCE and DWI images in the presence of Gaussian and Rician noise. Evaluation is done in both variable and fixed level of noise. Both subjective and objective quality assessment parameters are considered for determining the final rating of filters executed over 300 mp-MRI images. This study concludes that anisotropic and NLM filter should be opted for denoising task because of their structural and other crucial details preserving capability.

Detection algorithm for turbulent interfaces and large-scale structures in intermittent flows

A robust algorithm is introduced for detection of large-scale coherent structures in transitional and intermittent flows that feature turbulent/non-turbulent (T/NT) interfaces. The algorithm is applicable the instantaneous flow fields of wall-bounded and free shear flows, and can effectively identify coherent events in the velocity or vorticity fields, or sweep/ejection motions. A database from direct numerical simulation (DNS) of transitional boundary layer is used to develop and demonstrate the capabilities of the algorithm which consists of three steps. The first is identification of the T/NT interface by comparing the normalized vorticity magnitude to a threshold value that is independent of the Reynolds number. The vorticity normalization is specifically designed to be applicable in transitional flows, where regions of the flow can host juxtaposed regions of laminar and turbulent flow. With the definition of the T/NT interface, conditional statistics are computed and perturbation quantities are defined relative to their respective conditional means. Second, the influence of the small-scale turbulence is excluded by applying an anisotropic Gaussian filter. The filter size is determined from the spatial characteristics of the small-scale vortical motions. In the third step, one-dimensional cores and two-dimensional surfaces within the flow structures of interest are identified from local extrema in the fields, and are tracked as Lagrangian objects. Using the algorithm, the population trends and advection speeds of large-scale sweep/ejection events are computed in the transitional boundary layer. Two additional flow configurations are also considered: turbulent jet flow emerging from a circular nozzle and the turbulent flow in a channel with a wavy surface.