2D Median Filter Research Articles

A UNIFIED PARADIGM OF CLASSIFYING GI TRACT DISEASES IN ENDOSCOPY IMAGES USING MULTIPLE FEATURES FUSION

The automatic identification of Gastrointestinal (GI) tract diseases in endoscopy images has been associated with the domain of medical imaging and computer vision. Its classification has various challenges, including color, low contrast, lesion shape, and complex background. A Deep features-based method for the classification of gastrointestinal disease is implemented in this article. The method suggested involves four significant steps: preprocessing, extraction of handcrafted, and deep Convolutional neural network features (Deep CNN), selection of solid features, fusion, and classification. 3D-Median filtering in the preprocessing stage increases the lesion contrast. The second stage extracts the features centered on the shape. The extracted features are of three types: HOG features, ResNet50, and Xception. Principal Component Analysis (PCA) is chosen to select extracted features, combined by concatenating them in a single array. A support vector system eventually categorizes fused features into multiple classes. The Kvasir dataset is used for the proposed model. The SVM has outstanding efficiency reached 96.6 percent, showing the proposed system's robustness

An optical-digital method of noise suppression in digital holography

Digital holography is widely used for the fast and accurate reconstruction of objects and three-dimensional (3D)-scene characteristics, such as intensity, depth locations, density, temperature, refractive index, etc. There are various types of noise in digital holography, including speckle noise, zero order and twin image, camera shot noise, and fixed-pattern noise. Therefore, additional hologram processing is required to obtain a high-quality reconstructed image. This paper proposes an optical-digital method of noise suppression in digital holography. First, a set of uncorrelated holograms is registered. The reconstructed images are treated as a 3D array, to which a 3D filter is then applied. A 3D median filter is used in this study; however, other 3D filters based on weighted averages can also be utilized. The proposed method is verified using computer-generated and optically registered digital Fresnel holograms. Even with a small number of images, the method significantly improves the quality: the normalized standard deviation was decreased by up to 3–10 times and the speckle contrast was reduced by more than six times. Registration of only ten holograms is sufficient to obtain a quality close to the maximum. The flexibility of the proposed method allows the ratio of the reconstruction quality to be varied according to the processing time.

How separable median filters can get better results than full 2D versions

The well-known method of median filtering is used both in a wide range of application frameworks and as a standalone filter. Small-window median filters can highly reduce the power of salt and pepper or additive Gaussian noise and minimize edge blurring. Large-window filters are also used, for example to estimate the background of images. Currently, the window size should not be an issue as a constant time algorithm and several implementations, including GPU-based codes, have been proposed in recent years. Unfortunately, none of these constant time implementations manage to fully exploit the capabilities of modern GPUs, and thus the throughputs of large-window median filters remain far below the peak throughputs allowed by recent GPU models. This paper aims at showing that a separable approximation of a 2D median filtering is often stronger than its full 2D implementation. Statistical and theoretical analysis are conducted to show and explain this fact which had so far remained unobserved. It is confirmed by experimentations on a dataset composed of 10,000 images, corrupted by different levels of salt and pepper noise. Separable and full 2D median filter algorithms are compared with several metrics, notably PSNR and MSSIM. In addition, a GPU implementation of 2D separable median filters is also proposed. This implementation is able to output up to 125 billion pixels per second on a recent Volta V100, which significantly outperforms existing implementations like Nvidia’s NPP library or Green’s code, resulting in the fastest median filtering solution to date.

An Efficient Breast Cancer Detection Framework for Medical Diagnosis Applications

Breast cancer is the most common type of cancer, and it is the reason for cancer death toll in women in recent years. Early diagnosis is essential to handle breast cancer patients for treatment at the right time. Screening with mammography is the preferred examination for breast cancer, as it is available worldwide and inexpensive. Computer-Aided Detection (CAD) systems are used to analyze medical images to detect breast cancer, early. The death rate of cancer patients has decreased by detecting tumors early and having appropriate treatment after operations. Processing of mammogram images has four main steps: pre-processing, segmentation of the region of interest, feature extraction and classification of the images into normal or abnormal classes. This paper presents an efficient framework for processing of mammogram images and introduces an algorithm for segmentation of the images to detect masses. The pre-processing step of mammogram images includes removal of digitization noise using a 2D median filter, removal of artifacts using morphological operations, and contrast enhancement using a fuzzy enhancement technique. The proposed fuzzy image enhancement technique is analyzed and compared with conventional techniques based on an Enhancement Measure (EME) and local contrast metrics. The comparison shows an outstanding performance of the proposed technique from the visual and numerical perspectives. The segmentation process is performed using Otsu's multiple thresholding method. This method segments the image regions into five classes with variable intensities using four thresholds. Its effectiveness is measured based on visual quality of the segmentation output, as it gives details about the image and positions of masses. The performance of the proposed framework is measured using Dice coefficient, Hausdorff, and Peak Signal-to-Noise Ratio (PSNR) metrics. The segmented tumor region with the proposed segmentation method is 81% of the ground truth region provided by an expert. Hence, the proposed framework achieves promising results for aiding radiologists in screening of mammograms, accurately.

Improved Background Subtraction Technique for Detecting Moving Objects

Introduction: Moving object detection from videos is among the most difficult task in different areas of computer vision applications. Among the traditional object detection methods, researchers conclude that Background Subtraction method carried out better in aspects of execution time and output quality. Mehtod: Visual background extractor is a renowned algorithm in Background Subtraction method for detecting moving object in various applications. In the recent years, lots of work has been carried out to improve the existing Visual Background extractor algorithm. Result: After investigating many state of art techniques and finding out the research gaps, this paper presents an improved background subtraction technique based on morphological operation and 2D median filter for detecting moving object which reduces the noise in the output video and also enhances its accuracy at a very limited additional cost. Experimental results in several benchmark datasets confirmed the superiority of the proposed method over the state-of-the-art object detection methods. Conclusion: In this article, a method has been proposed for moving object detection where the quality of the output object is enhanced and good accuracy is achieved. This method provide with accurate experimental results, which helps in efficient object detection. The proposed technique also deals with Visual Background extractor Algorithm along with the Image Enhancement Procedure like Morphological and 2-D Filtering at a limited additional cost Discussion: This article worked on certain specific field, like noise reduction and image enhancement of output images of the existing ViBe Algorithm. The technique proposed in this article will be beneficial for various computer vision applications like video surveillance, road condition monitoring, airport safety, human activity analysis, monitoring marine border for security purpose etc.

Simultaneous external and internal inspection of a cylindrical CFRP lattice-skin structure based on rotational ultrasonic propagation imaging and laser displacement sensing

Composite lattice-skin structures have been widely used in the aerospace industry to reduce structure weight and fabrication cost. The filament winding technique can be used to manufacture them automatically. However, several defects can occur in every manufacturing step, such as outward dimples (in which the skin swells during the curing process) and de-bonding (in which the rib and skin detach). They drastically reduce the buckling strength under compression and should be evaluated before operation. In this study, we simultaneously inspected the outside and inside of a cylindrical composite lattice-skin structure and visualized external and internal defects based on laser displacement sensing and rotational ultrasonic propagation imaging. An external inspection algorithm that used a laser displacement sensor was added to a previously developed internal inspection algorithm. Then, the accuracy of the laser displacement sensor was confirmed by measuring the groove depth of the composite brake disk. Thereafter, external and internal inspection was simultaneously performed on the composite lattice-skin structure. This system can clearly visualize not only external defects (dimples) using laser displacement sensing with 2D median filtering and 2D fast-Fourier transform but also internal defects (de-bonding) using ultrasonic propagation imaging.

Establishing Normative Embryonic and Early Postnatal Cranial Bone Volume and Density Values for C57BL/6J Laboratory Mice

We provide normative data for cranial bone volume and tissue mineral density (TMD) for embryonic and early postnatal C57BL/6J mice that can be used as control data for researchers using experimental mouse models. C57BL/6J mice were selected based on their common usage in experimental research. Our sample includes 3 males and 3 females at four ages: embryonic day (E) 15.5, E17.5, postnatal day (P) 0, and P7. MicroCT images were acquired by the Center for Quantitative Imaging at the Pennsylvania State University using the General Electric v|tom|x L300 nano/microCT system. Prior to data collection, all postcranial bones were electronically removed from the scan using Avizo 9.4, and then background noise was removed from each scan using a 3D median filter. Isosurfaces were reconstructed to represent cranial bone based on a hydroxyapatite phantom that was scanned with each specimen. To maintain consistency between specimens and across age groups, we used 85 mg/cm3 as the minimum threshold to create isosurfaces and segment bone for this study. Twenty cranial bones were individually segmented for each specimen, and volume and TMD were estimated in Avizo. After correcting for multiple testing, no significant differences were found in bone volume or TMD between males and females for any of the examined cranial bones at any age, and data were pooled for analyses. As expected, volume of ossified bone increased with age from E15.5 to P7. The number of ossification centers present in a single individual is highly variable at E15.5 (ranging from 2 – 10), with all bones present by P7. Most of the measured bones consistently increased in TMD with age, with several exceptions. The palatine, vomer, and zygomatic bones each decreased in TMD from P0 to P7. The TMD of the maxilla also decreased slightly from E17.5 to P0, but then increased from P0 to P7. No significant patterns of development related to originating cell population (mesoderm v. neural crest) or mode of ossification (endochondral v. intramembranous) were apparent. These data can serve as a reference standard for future studies that examine craniofacial growth and disease models using mice bred on a C57BL/6J background.Support or Funding InformationNational Science Foundation Grant BCS‐1731909, NIDCR R01DE027677 and R01 DE022988, and the Pennsylvania State University Energy and Environmental Sustainability Laboratories Green Seed Grant

Accurate Breast Cancer Prediction using Machine Learning Techniques

Applications of machine learning (ML) have been increasing widely in various fields like recommendation, fault identification and disease prediction. In ML, different algorithms were available and utilized in disease prediction such as heart disease prediction, cancer prediction and other forms disease prediction. In our proposed work, breast cancer prediction using ML has been implemented. Initially breast cancer images have been taken as input, preprocessing steps will be done to remove noisy and irrelevant data from image. Then 2D median filter is a nonlinear operation often used in image processing to reduce "salt and pepper" noise. To increase contrast of image contrast-limited adaptive histogram equalization is used. Segmentation has been implemented and GLCM feature extraction is deployed based on this information classification is implemented. For accurate classification Artificial neural network (ANN) is used to predict whether the patient is affected by breast cancer or not. Compared to other existing method our method predicts results in accurate way.

Optimization Algorithm of UWB Positioning for Aircraft Assembly Workshop

Due to the complexity of strong metal interference and multiple occlusions in aircraft assembly workshop, the random 'drift' phenomenon often happens in the ultra wide band (UWB) based positioning system. To solve this, a fusion positioning optimization algorithm is proposed based on median filtering, hidden Markov model (HMM) and Kalman filtering. Firstly, based on the three-dimensional (3D) median filtering, a queue optimization method with weights is introduced to smooth the measurement data and eliminate the abnormal value. Secondly, taking Singer model as a reference, a single-dimension acceleration distribution model is designed. In order to further consider the spatial motion characteristics of objects in workshop, the distribution is extended from 1D to 3D, and discretized into the state quantity of HMM. Subsequently, the data obtained by the two methods are fused by taking Kalman filter as an iterator, and then the optimized location solution is obtained based on dynamic weights. Finally, an experiment is conducted in an aircraft assembly workshop. Results show that 99.3% of dynamic positioning errors are less than 15 cm after using the proposed algorithm. Even in the situation with large signal-fluctuation, there are 71.6% of positioning data whose errors are reduced. The random 'drift' is greatly decreased.

3D Object Recognition System Based On Local Shape Descriptors and Depth Data Analysis

Background: A physical object, which is actually in 3D form, is captured by a sensor/ camera (in case of computer vision) and seen by a human eye (in case of a human vision). When someone is observing something, many other things are also involved there which make it more challenging to recognize. After capturing such a thing by a camera or sensor, a digital image is formed which is nothing other than a bunch of pixels. It is becoming important to know that how a computer understands images. Objective: This paper is for highlighting novel techniques on 3D object recognition system with local shape descriptors and depth data analysis. Methods: The proposed work is applied to RGBD and COIL-100 datasets and this is of four-fold as preprocessing, feature generation, dimensionality reduction, and classification. The first stage of preprocessing is smoothing by 2D median filtering on the depth (Z-value) and registration by orientation correction on 3D object data. The next stage is of feature generation and having two phases of shape map generation with shape index map and SIFT/SURF descriptors. The dimensionality reduction is the third stage of this proposed work where linear discriminant analysis and principal component analysis are used. The final stage is fused on classification. Results: Here, calculation of the discriminative subspace for the training set, testing of object data and classification is done by comparing target and query data with different aspects for finding proper matching tasks. Conclusion: This concludes with new proposed approach of 3D Object Recognition. The local shape descriptors are used for 3D object recognition system to implement and test. This system is achieves 89.2% accuracy for Columbia object image library-100 images by using local shape descriptors.

A Low-Cost Design of 2D Median Filter

The median filter plays an important role in image processing for noise removing. To enhance the operating speed, hardware implementation of median filter on field programmable gate arrays (FPGA) or application specific IC (ASIC) are necessary and inevitable. This paper presents a low-cost and high-throughput hardware design for two-dimensional (2D) median filter. Two techniques are employed to reduce circuit area. One is the parallel three-valued sorting method for reducing the number of pipelined registers under constant speed, and the other is the functional sharing method used to implement median value sorting with the least number of comparisons. The proposed 2D median filter is implemented by two different synthesized methods. One is synthesized by Synopsys Design Compiler with the TSMC 90nm cell library (ASIC design) and the other is synthesized by Xilinx FPGA Virtex7 XC7VX690T (FPGA design). Experimental results show that the area cost of the proposed design is reduced more than 30% on average in comparison to the previous designs.

Automatic Identification of Blood Vessels, Exaudates and Abnormalities in Retinal Images for Diabetic Retinopathy Analysis

Now a days the diabetic retionopathy is become one of the serious type of disease which is very much responsible for blindness of eye. It is become a serious problems throughout allover the world. For some cases patients have lack of the consciousness of disease until it becomes so much late for providing the treatment properly. Early detection and treatment is become important for more prevention of probability of color blindness. The main objectives are to efficiently segmentation of blood vessels in retinal area and to detect and classify the areas of abnormalities present in eye. The work is based on different segmentation approach, edge detection and classification is done through SVM classifier. Here Image Preprocessing is basically performed with 2D median filter and histogram equalization for getting image in much enhanced form. Then extraction of features set as microaneurysm, hemorrhages, exudates, cotton wool spots and variations in blood vessels etc is performed in the effected retinal images. In this approach, some statistical parameter like as entropy, mean, RMS, correlation, variance, standard deviation are used for statistical analysis of diseases. The Technique uses standard datasets like as DRIONS, DIARECTDB0, DIARECTDB1for retinal image of working in this field. This techniques shows how easily we can detect region of abnormalities present in images with more than 80% of accuracy rates using SVM classifier.

An Optimized Method for Segmentation and Classification of Apple Diseases Based on Strong Correlation and Genetic Algorithm Based Feature Selection

Agriculture is a major part of the world economy as it provides food safety. However, in recent years, it has been noted that plants are extensively infected by different diseases. This causes enormous economic losses in agriculture industry around the world. The manual inspection of fruit diseases is a difficult process which can be minimized by using automated methods for detection of plant diseases at the earlier stage. In this paper, a new method is implemented for apple diseases identification and recognition. Three pipeline procedures are followed by preprocessing, spot segmentation, and features extraction, and classification. In the first step, the apple leaf spots are enhanced by a hybrid method which is the conjunction of 3D box filtering, de-correlation, 3D-Gaussian filter, and 3D-median filter. After that, the lesion spots are segmented by the strong correlation-based method and optimized their results by fusion of expectation maximization (EM) segmentation. Finally, the color, color histogram, and local binary pattern (LBP) features are fused by comparison-based parallel fusion. The extracted features are optimized by genetic algorithm and classified by One-vs-All M-SVM. The experimental results are performed on plant village dataset. The proposed methodology is tested for four types of apple disease classes including healthy leaves, Blackrot, Rust, and Scab. The classification accuracy shows the improvement of our method on selected apple diseases. Moreover, the good preprocessing step always produced prominent features which later achieved significant classification accuracy.

AUTOMATED ULCER AND BLEEDING CLASSIFICATION FROM WCE IMAGES USING MULTIPLE FEATURES FUSION AND SELECTION

In the area of medical imaging and computer vision, automatic diagnosis of ulcer and bleeding from wireless capsule endoscopy images has been an active research domain. It contains several challenges including low contrast, complex background, lesion shape and color which affect its segmentation and classification accuracy. In this article, a novel method for automated detection and classification of stomach infection is implemented. The proposed method consists of four major steps including preprocessing, lesion segmentation, image representation and classification. The lesion contrast is improved in preprocessing step by employing 3D-box filtering, 3D-median filtering and HSV transformation. In the second step, geometric features are extracted and applied to the saturated channel to give a binary image. The binary image is further improved by fusion of generated mask. After that, extraction of three types of features including color, shape and surf is performed from HSV and binary segmented images and their information is fused by a serial based method. A principal component analysis (PCA) and correlation coefficient based feature selection approach is proposed which are classified by multi class support vector machine (M-SVM). The proposed method is evaluated on personally collected images of three different classes including ulcer, bleeding and healthy. The M-SVM performs well with a maximum accuracy of 98.3% which shows the authenticity of presented method.

Video oriented filter for impulse noise reduction

A window-adaptive video filter for removal of impulse noise from grayscale videos is proposed. The new method is based on local orientation estimation. The dominant orientation of the pattern in a local spatial neighborhood is computed by minimizing an expression of directional derivatives, and at the same time the orientation strength is also computed. Based on the local spatial orientation and its strength, the size, shape, and orientation of 3D filter window are adaptively determined, which leads to the proposed window-adaptive 3D median filter. To further enhance denoising performance, a new noise detection mechanism is developed and integrated to the proposed video filter. By using this noise detector, video pixels are classified into noise-free and noisy ones. For the noisy pixels detected, the proposed window-adaptive 3D filter is performed. Experimental results show that the proposed method outperforms other state-of-the-art video denoising methods in both objective measure and visual evaluation.

Detection of breast abnormalities in digital mammograms using the electromagnetism-like algorithm

The detection of abnormalities in the breast at an early stage can be so helpful for breast cancer treatment. Currently, mammography is the cheapest and the most efficient technique in terms of identifying the suspicious lesions in the breast. However, the interpretation of this screening remains so hard and could lead to inaccurate detection known as false positive and false negative. Dense breast mammograms particularly are difficult to read because they may contain abnormal structures that are similar to the normal breast tissue. In this paper, we introduce an effective method for the detection of the ambiguous areas in digital mammograms. After noise and artifacts removal from the images using 2D Median filtering and labeling, we isolate the abnormalities using the meta-heuristic algorithm Electromagnetism-like Optimization (EML). The segmentation was carried on the abnormal cases of two different databases Mini-Mias and DDSM. The accuracy detection rate achieves almost 85% for both databases and 91.07% for DDSM alone.

Object Segmentation from Background of 2D Image

One of the difficult tasks in the image processing field and still not solved is segmentation of object from background of 2D image accurately. Therefore, a new method has been proposed for the purpose of segmenting the object from its background for the purpose of enhancing the images and obtains characteristics of the object without the rest of the region of the image. This process is important to provide optimal classification in the process of pattern recognition. Therefore, this paper proposed the method that includes several tasks, after loading the six files of images; this work applies the segmentation al- gorithm depending on the border and the color of the object. Finally, 2D median filtering algorithm was employed to remove noisy objects of various shapes and sizes. The algo- rithm was tested on variety images, and the results are high precision. In the other words, the proposed method is able to segment the objects from the background with promising results.

Patient-specific Monte Carlo dose calculations for 103Pd breast brachytherapy

This work retrospectively investigates patient-specific Monte Carlo (MC) dose calculations for 103Pd permanent implant breast brachytherapy, exploring various necessary assumptions for deriving virtual patient models: post-implant CT image metallic artifact reduction (MAR), tissue assignment schemes (TAS), and elemental tissue compositions. Three MAR methods (thresholding, 3D median filter, virtual sinogram) are applied to CT images; resulting images are compared to each other and to uncorrected images. Virtual patient models are then derived by application of different TAS ranging from TG-186 basic recommendations (mixed adipose and gland tissue at uniform literature-derived density) to detailed schemes (segmented adipose and gland with CT-derived densities). For detailed schemes, alternate mass density segmentation thresholds between adipose and gland are considered. Several literature-derived elemental compositions for adipose, gland and skin are compared. MC models derived from uncorrected CT images can yield large errors in dose calculations especially when used with detailed TAS. Differences in MAR method result in large differences in local doses when variations in CT number cause differences in tissue assignment. Between different MAR models (same TAS), PTV and skin each vary by up to 6%. Basic TAS (mixed adipose/gland tissue) generally yield higher dose metrics than detailed segmented schemes: PTV and skin are higher by up to 13% and 9% respectively. Employing alternate adipose, gland and skin elemental compositions can cause variations in PTV of up to 11% and skin of up to 30%. Overall, AAPM TG-43 overestimates dose to the PTV ( on average 10% and up to 27%) and underestimates dose to the skin ( on average 29% and up to 48%) compared to the various MC models derived using the post-MAR CT images studied herein. The considerable differences between TG-43 and MC models underline the importance of patient-specific MC dose calculations for permanent implant breast brachytherapy. Further, the sensitivity of these MC dose calculations due to necessary assumptions illustrates the importance of developing a consensus modelling approach.

Development of virtual patient models for permanent implant brachytherapy Monte Carlo dose calculations: interdependence of CT image artifact mitigation and tissue assignment

This work investigates and compares CT image metallic artifact reduction (MAR) methods and tissue assignment schemes (TAS) for the development of virtual patient models for permanent implant brachytherapy Monte Carlo (MC) dose calculations. Four MAR techniques are investigated to mitigate seed artifacts from post-implant CT images of a homogeneous phantom and eight prostate patients: a raw sinogram approach using the original CT scanner data and three methods (simple threshold replacement (STR), 3D median filter, and virtual sinogram) requiring only the reconstructed CT image. Virtual patient models are developed using six TAS ranging from the AAPM-ESTRO-ABG TG-186 basic approach of assigning uniform density tissues (resulting in a model not dependent on MAR) to more complex models assigning prostate, calcification, and mixtures of prostate and calcification using CT-derived densities. The EGSnrc user-code BrachyDose is employed to calculate dose distributions. All four MAR methods eliminate bright seed spot artifacts, and the image-based methods provide comparable mitigation of artifacts compared with the raw sinogram approach. However, each MAR technique has limitations: STR is unable to mitigate low CT number artifacts, the median filter blurs the image which challenges the preservation of tissue heterogeneities, and both sinogram approaches introduce new streaks. Large local dose differences are generally due to differences in voxel tissue-type rather than mass density. The largest differences in target dose metrics (D90, V100, V150), over 50% lower compared to the other models, are when uncorrected CT images are used with TAS that consider calcifications. Metrics found using models which include calcifications are generally a few percent lower than prostate-only models. Generally, metrics from any MAR method and any TAS which considers calcifications agree within 6%. Overall, the studied MAR methods and TAS show promise for further retrospective MC dose calculation studies for various permanent implant brachytherapy treatments.

An Automatic Segmentation for Determination of IV Vessel Boundaries

Coronary heart disease ranks as the top disease which causes the death in the Egypt. Intravascular (IV) ultrasound imaging is used along with X-ray coronary angiography to detect vessel pathologies. This paper introduces three automated approaches for detection of lumen and media-adventitia borders in IV images. In the first method, a 2D median filter is used to initialize the lumen and media-adventitia contours. The second method uses the Lucy-Richardson method and the 2D median filter to determine the lumen border. Method three is a combination of two methods to combine their advantages. The catheter boundaries were automatically determined. This method shows a higher accuracy and ability to work under artifacts and noisy conditions.