Showing papers in "Signal, Image and Video Processing in 2014"
TL;DR: Feed-forward back-propagation neural network has been used for classification and training algorithm for this network that updates the weight and bias values according to Levenberg–Marquardt optimization technique to detect seizures with 100 % classification accuracy using artificial neural network.
Abstract: There are numerous neurological disorders such as dementia, headache, traumatic brain injuries, stroke, and epilepsy. Out of these epilepsy is the most prevalent neurological disorder in the human after stroke. Electroencephalogram (EEG) contains valuable information related to different physiological state of the brain. A scheme is presented for detecting epileptic seizures from EEG data recorded from normal subjects and epileptic patients. The scheme is based on discrete wavelet transform (DWT) analysis and approximate entropy (ApEn) of EEG signals. Seizure detection is performed in two stages. In the first stage, EEG signals are decomposed by DWT to calculate approximation and detail coefficients. In the second stage, ApEn values of the approximation and detail coefficients are calculated. Significant differences have been found between the ApEn values of the epileptic and the normal EEG allowing us to detect seizures with 100 % classification accuracy using artificial neural network. The analysis results depicted that during seizure activity, EEG had lower ApEn values compared to normal EEG. This gives that epileptic EEG is more predictable or less complex than the normal EEG. In this study, feed-forward back-propagation neural network has been used for classification and training algorithm for this network that updates the weight and bias values according to Levenberg–Marquardt optimization technique.
224 citations
TL;DR: The algorithms for the detection and removal of rain from videos have been reviewed and merits and demerits of the algorithms are discussed, which motivate the further research.
Abstract: In this paper, the algorithms for the detection and removal of rain from videos have been reviewed. Rain reduces the visibility of scene and thus performance of computer vision algorithms which use feature information. Detection and removal of rain requires the discrimination of rain and nonrain pixels. Accuracy of the algorithm depends upon this discrimination. Here merits and demerits of the algorithms are discussed, which motivate the further research. A rain removal algorithm has a wide application in tracking and navigation, consumer electronics and entertainment industries.
100 citations
TL;DR: A novel vision-based fall detection method for monitoring elderly people in house care environment using ellipse fitting and an integrated normalized motion energy image computed over a short-term video sequence is proposed.
Abstract: Fall detection is one of the most important health care issues for elderly people at home, which can lead to severe injuries. With the advances and conveniences in computer vision in the last few decades, computer vision-based methods provide a promising way for detecting falls. In this paper, we propose a novel vision-based fall detection method for monitoring elderly people in house care environment. The foreground human silhouette is extracted via background modeling and tracked throughout the video sequence. The human body is represented with ellipse fitting, and the silhouette motion is modeled by an integrated normalized motion energy image computed over a short-term video sequence. Then, the shape deformation quantified from the fitted silhouettes is used as the features to distinguish different postures of the human. Finally, different postures are classified via a multi-class support vector machine and a context-free grammar-based method that provides longer range temporal constraints can verify the detected falls. Extensive experiments show that the proposed method has achieved a reliable result compared with other common methods.
92 citations
TL;DR: Some future considerations that can be applied in this topic such as: the fusion between different techniques previously used, use both ECG and PCG signals in a multimodal biometric authentication system and building a prototype system for real-time authentication.
Abstract: Due to the great advances in biomedical digital signal processing, new biometric traits have showed noticeable improvements in authentication systems. Recently, the ElectroCardioGram (ECG) and the PhonoCardioGraph (PCG) have been proposed as novel biometrics. This paper aims to review the previous studies related to the usage of the ECG and PCG signals in human recognition. In addition, we discuss briefly the most important techniques and methodologies used by researchers in the preprocessing, feature extraction and classification of the ECG and PCG signals. At the end, we introduce some future considerations that can be applied in this topic such as: the fusion between different techniques previously used, use both ECG and PCG signals in a multimodal biometric authentication system and building a prototype system for real-time authentication.
78 citations
TL;DR: Based on a new convolution operation, convolution and correlation theorems are formulated for the offset linear canonical transform and the convolution theorem is used to investigate the sampling theorem for the band-limited signal in the OLCT domain.
Abstract: The offset linear canonical transform (OLCT), which is a time-shifted and frequency-modulated version of the linear canonical transform, has been shown to be a powerful tool for signal processing and optics. However, some basic results for this transform, such as convolution and correlation theorems, remain unknown. In this paper, based on a new convolution operation, we formulate convolution and correlation theorems for the OLCT. Moreover, we use the convolution theorem to investigate the sampling theorem for the band-limited signal in the OLCT domain. The formulas of uniform sampling and low-pass reconstruction related to the OLCT are obtained. We also discuss the design method of the multiplicative filter in the OLCT domain. Based on the model of the multiplicative filter in the OLCT domain, a practical method to achieve multiplicative filtering through convolution in the time domain is proposed.
70 citations
TL;DR: This paper first proposes a generalized convolution theorem for the LCT and then derives a corresponding product theorem associated with the L CT, which is shown to be special cases of the derived results.
Abstract: The linear canonical transform (LCT), which is a generalized form of the classical Fourier transform (FT), the fractional Fourier transform (FRFT), and other transforms, has been shown to be a powerful tool in optics and signal processing. Many results of this transform are already known, including its convolution theorem. However, the formulation of the convolution theorem for the LCT has been developed differently and is still not having a widely accepted closed-form expression. In this paper, we first propose a generalized convolution theorem for the LCT and then derive a corresponding product theorem associated with the LCT. The ordinary convolution theorem for the FT, the fractional convolution theorem for the FRFT, and some existing convolution theorems for the LCT are shown to be special cases of the derived results. Moreover, some applications of the derived results are presented.
60 citations
TL;DR: A segmentation technique that combines the features of fuzzy c-mean (FCM) clustering and region-based active contour method that has advantages in accuracy in comparison with standard region growing method and FCM for the detection of hemorrhage from brain CT images is presented.
Abstract: Intracranial hemorrhage (ICH) detection is the primary task for the patients suffering from neurological disturbances and head injury. This paper presents a segmentation technique that combines the features of fuzzy c-mean (FCM) clustering and region-based active contour method. In the suggested method, the fuzzy membership degree from FCM clustering is first used to initialize the active contour, which propagates for the detection of the desired object. In addition to active contour initialization, the fuzzy clustering is also used to estimate the contour propagation controlling parameters. The level set function as used by active contour in the proposed method does not need re-initialization process; thus, it fastens the convergent speed of the contour propagation. The efficacy of the suggested method is demonstrated on a dataset of 20 brain computed tomography (CT) images suffered with ICH. Experimental results show that the proposed method has advantages in accuracy in comparison with standard region growing method and FCM for the detection of hemorrhage from brain CT images.
60 citations
TL;DR: The proposed DSR-SVD technique is found to give noteworthy better performance in terms of contrast enhancement factor, color enhancement factor and perceptual quality measure.
Abstract: In this paper, a dynamic stochastic resonance (DSR)-based technique in singular value domain for contrast enhancement of dark images has been presented. The internal noise due to the lack of illumination is utilized using a DSR iterative process to obtain enhancement in contrast, colorfulness as well as perceptual quality. DSR is a phenomenon that has been strategically induced and exploited and has been found to give remarkable response when applied on the singular values of a dark low-contrast image. When an image is represented as a summation of image layers comprising of eigen vectors and values, the singular values denote luminance information of each such image layer. By application of DSR on the singular values using the analogy of a bistable double-well potential model, each of the singular values is scaled to produce an image with enhanced contrast as well as visual quality. When compared with performance of some existing spatial domain enhancement techniques, the proposed DSR-SVD technique is found to give noteworthy better performance in terms of contrast enhancement factor, color enhancement factor and perceptual quality measure.
59 citations
TL;DR: A low-complexity data-adaptive PVC recognition approach with good robustness against noise and generalization capability is developed, demonstrating a good potential in real-time application.
Abstract: Premature ventricular contraction (PVC) may lead to life-threatening cardiac conditions. Real-time automated PVC recognition approaches provide clinicians the useful tools for timely diagnosis if dangerous conditions surface in their patients. Based on the morphological differences of the PVC beats in the ventricular depolarization phase (QRS complex) and repolarization phase (mainly T-wave), two beat-to-beat template-matching procedures were implemented to identify them. Both templates were obtained by a probability-based approach and hence were fully data-adaptive. A PVC recognizer was then established by analyzing the correlation coefficients from the two template-matching procedures. Our approach was trained on 22 ECG recordings from the MIT-BIH arrhythmia database (MIT-BIH-AR) and then tested on another 22 nonoverlapping recordings from the same database. The PVC recognition accuracy was 98.2 %, with the sensitivity and positive predictivity of 93.1 and 81.4 %, respectively. To evaluate its robustness against noise, our approach was applied again to the above testing set, but this time, the ECGs were not preprocessed. A comparable performance was still obtained. A good generalization capability was also confirmed by validating our approach on an independent St. Petersburg Institute of Cardiological Technics database. In addition, our performance was comparable with these published complex approaches. In conclusion, we have developed a low-complexity data-adaptive PVC recognition approach with good robustness against noise and generalization capability. Its performance is comparable to other state-of-the-art methods, demonstrating a good potential in real-time application.
54 citations
TL;DR: A new fuzzy weighting function is introduced, which can shut off the impulsive weight effectively, to the non-local means, and the more a pixel is corrupted, the less it is exploited to reconstruct image information.
Abstract: In this paper, we propose a fuzzy weighted non-local means filter for the removal of random-valued impulse noise. We introduce a new fuzzy weighting function, which can shut off the impulsive weight effectively, to the non-local means. According to the new weighting function, the more a pixel is corrupted, the less it is exploited to reconstruct image information. Experiments show that the performances of the new filter are surprisingly satisfactory in terms of both visual quality and quantitative measurement. Moreover, our filter also can be used to remove mixed Gaussian and random-valued impulse noise.
50 citations
TL;DR: The proposed enhancement technique using the fuzzy set theory for low contrast and nonuniform illumination images exhibits the best performance and defeats other methods in terms of preserving brightness and details without amplifying existing noises.
Abstract: This paper presents a new enhancement technique using the fuzzy set theory for low contrast and nonuniform illumination images. A new parameter called the contrast factor which will provide information on the difference among the gray-level values in the local neighborhood is proposed. The contrast factor is measured by both local and global information to ensure that the fine details of the degraded image are enhanced. This parameter is used to divide the degraded image into bright and dark regions. The enhancement process is applied on gray-scale images wherein the modified Gaussian membership function is employed. The process is performed separately according to the image’s respective regions. The performance of the proposed method is comparable with other state-of-the-art techniques in terms of processing time. The proposed method exhibits the best performance and defeats other methods in terms of preserving brightness and details without amplifying existing noises.
TL;DR: The experimental results show that the proposed hybrid watermarking method provides robustness against several geometric distortions, signal processing operations, combined distortions and photo editing.
Abstract: In this paper, we present a robust hybrid watermarking method applied to color images for authentication, which presents robustness against several distortions. Due to the different nature of common signal processing and geometrical attacks, two different techniques for embed a same watermark are used in this method. In the first one, the luminance component (Y) information is used to embed the watermark bit sequence into the magnitude of the middle frequencies of the Discrete Fourier Transform (DFT). In the second one, a selected region of 2D histogram composed by blue-difference and red-difference (Cb–Cr) chrominance components is modified according to the watermark bit sequence. The quality of the watermarked image is measured using the following well-known indices peak signal to noise ratio (PSNR), visual information fidelity (VIF) and structural similarity index (SSIM). The difference color of the watermarked image is obtained using the normalized color difference (NCD) measure. The experimental results show that the proposed method provides robustness against several geometric distortions, signal processing operations, combined distortions and photo editing. The comparison with the previously reported methods based on different techniques is also provided.
TL;DR: A novel random forest (RF)-based segmentation and classification approaches for the automated classification of Mycobacterium tuberculosis in microscopic images of Ziehl–Neelsen-stained sputum smears obtained using a light-field microscope is proposed.
Abstract: The World Health Organization suggests visual examination of stained sputum smear samples as a preliminary and basic diagnostic technique for diagnosing tuberculosis. The visual examination process requires much time of laboratorian, and also, it is prone to mistakes. For this purpose, this paper proposes a novel random forest (RF)-based segmentation and classification approaches for the automated classification of Mycobacterium tuberculosis in microscopic images of Ziehl–Neelsen-stained sputum smears obtained using a light-field microscope. The RF supervised learning method is improved to classify each pixel depending on local color distributions as a part of candidate bacilli regions. Therefore, each pixel is labeled as either a candidate tuberculosis (TB) bacilli pixel or not. The candidate pixels are grouped together using connected component analysis. Each pixel group is then rotated, resized and centrally positioned within a bounding box, respectively, in order to utilize appearance-based tuberculosis bacteria identification algorithms. Finally, each region is classified by using the proposed RF learning algorithm trained on manually marked TB bacteria regions in the training images. The algorithm produces results that agree well with manual segmentation and identification. Different two-class pixel and object classifiers are also compared to show the performance of the proposed RF-based pixel segmentation and bacilli objects identification algorithm. The sensitivity and specificity of the proposed classifier are above 75.77 and 96.97 % for the segmentation of the pixels, respectively. It is also revealed that the sensitivity increases over 93 % when the staining is performed in accordance with the procedure. Moreover, these measures are above 89.34 and 62.89 % for the identification of bacilli objects. The results show that the proposed novel method is quite successful when compared to the other applied methods.
TL;DR: The experimental results obtained of the MIT-BIH Arrhythmia database showed that for all feature representation adopted in this work, the GP detector trained only with 600 beats from PVC and Non-PVC classes can provide an overall accuracy and a sensitivity above 90 % on 20 records.
Abstract: In this paper, we propose to investigate the capabilities of two kernel methods for the detection and classification of premature ventricular contractions (PVC) arrhythmias in Electrocardiogram (ECG signals). These kernel methods are the support vector machine and Gaussian process (GP). We propose to study these two classifiers with various feature representations of ECG signals, such as morphology, discrete wavelet transform, higher-order statistics, and S transform. The experimental results obtained on 48 records (i.e., 109,887 beats) of the MIT-BIH Arrhythmia database showed that for all feature representation adopted in this work, the GP detector trained only with 600 beats from PVC and Non-PVC classes can provide an overall accuracy and a sensitivity above 90 % on 20 records (i.e., 49,774 beats) and 28 records (i.e., 60,113 beats) seen and unseen, respectively, during the training phase.
TL;DR: Improved recognition accuracies are achieved compared to the individual systems and multimodal systems using other local or global feature extractors for both modalities.
Abstract: Fusion of multiple biometrics combines the strengths of unimodal biometrics to achieve improved recognition accuracy. In this study, face and iris biometrics are used to obtain a robust recognition system by using several feature extractors, score normalization and fusion techniques. Global and local feature extractors are used to extract face and iris features separately, and then, the fusion of these modalities is performed on different subsets of face and iris image databases of ORL, FERET, CASIA and UBIRIS. The proposed method uses Local Binary Patterns local feature extractor and subspace Linear Discriminant Analysis global feature extractor on face and iris images, respectively. Face and iris scores are normalized using tanh normalization, and then, Weighted Sum Rule is applied for the fusion of these two modalities. Improved recognition accuracies are achieved compared to the individual systems and multimodal systems using other local or global feature extractors for both modalities.
TL;DR: This algorithm has been found to be able to remove the high density salt-and-pepper noise and also preserved the fine details of the four images, Lena, Elaine, Rhythm, and Sunny, used as test images in this study.
Abstract: A new decision-based algorithm has been proposed for the restoration of digital images which are highly contaminated by the saturated impulse noise (i.e., salt-and-pepper noise). The proposed denoising algorithm performs filtering operation only to the corrupted pixels in the image, keeping uncorrupted pixels intact. The present study has used a coupled window scheme for the removal of high density noise. It has used sliding window of increasing dimension, centered at any pixel and replaced the noisy pixels consecutively by the median value of the window. However, if the entire pixels in the window are noisy, then the dimension of sliding window is increased in order to obtain the noise-free pixels for median calculation. Consequently, this algorithm has been found to be able to remove the high density salt-and-pepper noise and also preserved the fine details of the four images, Lena, Elaine, Rhythm, and Sunny, used as test images in this study (The latter two real-life images have been acquired using Sony: Steady Shot DSC- S3000). Experimentally, it has been found that the proposed algorithm yields better peak signal-to-noise ratio, image enhancement factor, structural similarity index measure and image quality index, compared with the other state-of-art median-based filters viz. standard median filter, adaptive median filter, progressive switched median filter, modified decision-based algorithm and modified decision-based unsymmetric trimmed median filter.
TL;DR: A new hybrid color image segmentation approach, which attempts two different transforms for texture representation and extraction, and the 2-D discrete wavelet transform and the contourlet transform that represents boundaries even more accurately are applied.
Abstract: This paper presents a new hybrid color image segmentation approach, which attempts two different transforms for texture representation and extraction. The 2-D discrete wavelet transform that can express the variance in frequency and direction of textures, and the contourlet transform that represents boundaries even more accurately are applied in our algorithm. The whole segmentation algorithm contains three stages. First, an adaptive color quantization scheme is utilized to obtain a coarse image representation. Then, the tiny regions are combined based on color information. Third, the proposed energy transform function is used as a criterion for image segmentation. The motivation of the proposed method is to obtain the complete and significant objects in the image. Ultimately, according to our experiments on the Berkeley segmentation database, our techniques have more reasonable and robust results than other two widely adopted image segmentation algorithms, and our method with contourlet transform has better performance than wavelet transform.
TL;DR: The proposed recursive spline interpolation filter is based on the neighborhood noise- free pixels and previous noise-free output pixel; hence, it is termed as recursive splines interpolationfilter.
Abstract: Spline-based approach is proposed to remove very high density salt-and-pepper noise in grayscale and color images. The algorithm consists of two stages, the first stage detects whether the pixel is noisy or noise-free. The second stage removes the noisy pixel by recursive spline interpolation filter. The proposed recursive spline interpolation filter is based on the neighborhood noise-free pixels and previous noise-free output pixel; hence, it is termed as recursive spline interpolation filter. The performance of the proposed algorithm is compared with the existing algorithms like standard median filter, decision-based filter, progressive switched median filter, and modified decision-based unsymmetric trimmed median filter at very high noise density. The proposed algorithm gives better peak signal-to-noise ratio, image enhancement factor, and correlation factor results than the existing algorithms.
TL;DR: It can be concluded from the experimental analysis that the performance of LDP along with ZMs is better than that of Z Ms alone and of ZMs along with other variants of LBP and LDP.
Abstract: Shape, being an important part of an object, has a special place in the field of shape-based image retrieval (SBIR). To retrieve most appropriate images, various descriptors are applied in SBIR like Zernike moments (ZMs), complex Zernike moments (CZMs) etc. Though ZMs/CZMs are good in SBIR but they are capable of extracting only global details of an image, hence something in addition to this is desirable to improve the performance of SBIR system. This paper presents experimental analysis of pixel-based dense descriptors such as local binary pattern (LBP), local directional pattern (LDP) and their variants. These descriptors are used as local features along with ZMs global features in achieving higher and accurate retrieval rate in SBIR system. We have analyzed these variants of LBP/LDP with various similarity measures on images. In case of ZMs, the magnitude component is used as global features. These methods are tested separately on suitable shape databases. Various databases used in the paper are MPEG-7 CE-2 region-based database, MPEG-7 CE-1 contour-based database and Trademark database. It can be concluded from the experimental analysis that the performance of LDP along with ZMs is better than that of ZMs alone and of ZMs along with other variants of LBP and LDP.
TL;DR: A novel center-symmetric scale invariant local ternary pattern feature is put forward to combine with pattern kernel density estimation for building a pixel-level-based background model, which is then used to detect moving foreground objects on every newly captured frame.
Abstract: This paper presents a real-time surveillance system for detecting and tracking people, which takes full advantage of local texture patterns, under a stationary monocular camera. A novel center-symmetric scale invariant local ternary pattern feature is put forward to combine with pattern kernel density estimation for building a pixel-level-based background model. The background model is then used to detect moving foreground objects on every newly captured frame. A variant of a fast human detector that utilizes local texture patterns is adopted to look for human objects from the foreground regions, and it is assisted by a head detector, which is proposed to find in advance the candidate locations of human, to reduce computational costs. Each human object is given a unique identity and is represented by a spatio-color-texture object model. The real-time performance of tracking is achieved by a fast mean-shift algorithm coupled with several efficient occlusion-handling techniques. Experiments on challenging video sequences show that the proposed surveillance system can run in real-time and is quite robust in segmenting and tracking people in complex environments that include appearance changes, abrupt motion, occlusions, illumination variations and clutter.
TL;DR: A selection method that can be applied to choose the best parameters to classify contractions in the uterine electrohysterography (EHG) signal for the detection of preterm labor indicates a set of 13 linear parameters, 3 nonlinear parameters and 2 propagation parameters that are potentially most useful to discriminate between pregnancy and labor contractions.
Abstract: This article proposes a selection method that can be applied to choose the best parameters to classify contractions in the uterine electrohysterography (EHG) signal for the detection of preterm labor. Several types of parameters have historically been extracted from the electrohysterogram. These can be divided into three classes: linear parameters, nonlinear parameters and parameters related to the electrohyterogram propagation. Frequency band enhancement EHG characterization has also been extensively studied. Our work is divided in two parts. The first part is to implement and compute all the parameters already extracted from the EHG that have been published in the literature. These parameters were computed both on the original EHG and on different frequency bands obtained using wavelet packet decomposition. In the second part, we will use a new parameters selection method to eliminate all parameters that are not efficient and pertinent for classification. Our results indicate a set of 13 linear parameters, 3 nonlinear parameters and 2 propagation parameters that are potentially most useful to discriminate between pregnancy and labor contractions, either on different frequency bands or directly on original EHG.
TL;DR: An attempt has been made to design a low-pass linear-phase multiplier-less finite duration impulse response (FIR) filter using differential evolution (DE) algorithm to search the impulse response coefficients of the FIR filter in the form of sum of power of two (SPT) in order to avoid the multipliers during design process.
Abstract: Evolutionary computational techniques have been employed judiciously in various signal processing applications of late. In this paper, such an attempt has been made to design a low-pass linear-phase multiplier-less finite duration impulse response (FIR) filter using differential evolution (DE) algorithm. This particular evolutionary optimization technique has been explored to search the impulse response coefficients of the FIR filter in the form of sum of power of two (SPT) in order to avoid the multipliers during design process. The performance of the designed low-pass filter has been studied thoroughly in terms of its frequency characteristics and primitive requirement of fundamental hardware blocks. The superiority of our design has been ascertained over a number of existing techniques by various means. Finally, the proposed filter of different lengths has been implemented on a field programmable gate array (FPGA) chip for evaluating the competency of this work. The percentage improvement in hardware complexity produced by our design has also been computed and clearly listed in this paper for convenience.
TL;DR: This study reports a fuzzy system to semantic image classification that is optimized by genetic algorithm after empirical design stage to materialize a robust rotation/lighting condition and size invariant image classifier.
Abstract: Image classification is a challenging problem of computer vision. This study reports a fuzzy system to semantic image classification. As it is a complex task, various information of digital image, including three color space components and two Zernike moments with different order, are gathered and utilized as an input of fuzzy inference system to materialize a robust rotation/lighting condition and size invariant image classifier. For better performance, all the membership functions are optimized by genetic algorithm after empirical design stage. 90.62 and 96.25 % classification rates for RGB and HSI color spaces confirm the reliability of optimized system in different image conditions given in this contribution.
TL;DR: This article has applied spatial fuzzy genetic algorithm (SFGA) for the unsupervised segmentation of color images and shows that the proposed technique outperforms state-of-the-art methods.
Abstract: Image segmentation is a very important low-level vision task. It is the perceptual grouping of pixels based on some similarity criteria. In this article, we have applied spatial fuzzy genetic algorithm (SFGA) for the unsupervised segmentation of color images. The SFGA adds diversity to the search process to find the global optima. The performance of SFGA is influenced by two factors: first, K number of clusters—should be known in advance; second, the initialization of the cluster centers. To overcome these issues, a progressive technique based on self-organizing map is presented to find out the optimal K number of clusters automatically. To handle the initialization problem, peaks are identified using the image color histograms. The genetic algorithm with fuzzy behavior maximizes the fuzzy separation and minimizes the global compactness among the segments. The segmentation is performed on wavelet transform image which not only reduces the dimensionality and computational cost but also makes more compact segments. A novel pruning technique is proposed to handle the problem of over-segmentation. The results show that the proposed technique outperforms state-of-the-art methods.
TL;DR: An efficient image denoising method that adaptively combines the features of wavelets, wave atoms and curvelets is presented, which results in perfect presentation of the smooth regions and efficient preservation of textures and edges in the image.
Abstract: This paper presents an efficient image denoising method that adaptively combines the features of wavelets, wave atoms and curvelets. Wavelet shrinkage is used to denoise the smooth regions in the image while wave atoms are employed to denoise the textures, and the edges will take advantage of curvelet denoising. The received noisy image is firstly decomposed into a homogenous (smooth/cartoon) part and a textural part. The cartoon part of the noisy image is denoised using wavelet transform, and the texture part of the noisy image is denoised using wave atoms. The two denoised images are then fused adaptively. For adaptive fusion, different weights are chosen from the variance map of the denoised texture image. Further improvement in denoising results is achieved by denoising the edges through curvelet transform. The information about edge location is gathered from the variance map of denoised cartoon image. The denoised image results in perfect presentation of the smooth regions and efficient preservation of textures and edges in the image.
TL;DR: The results based on recognition performance and ROC analysis demonstrate that the proposed matching score level fusion scheme using Weighted Sum rule, tanh normalization, iris-FVF and facial features extracted by LBP achieves a significant improvement over unimodal and multimodal methods.
Abstract: Multimodal biometrics-based systems aim to improve the recognition accuracy of human beings using more than one physical and/or behavioral characteristics of a person. In this paper, different fusion schemes at matching score level and feature level are employed to obtain a robust recognition system using several standard feature extractors. The proposed method involves the consideration of a face–iris multimodal biometric system using score level and feature level fusion. Principal Component Analysis (PCA), subspace Linear Discriminant Analysis (LDA), subpattern-based PCA, modular PCA and Local Binary Patterns (LBP) are global and local feature extraction methods applied on face and iris images. In fact, different feature sets obtained from five local and global feature extraction methods for unimodal iris biometric system are concatenated at feature level fusion called iris feature vector fusion (iris-FVF), while for unimodal face biometric system, LBP is used to achieve efficient texture descriptors. Feature selection is performed using Particle Swarm Optimization (PSO) at feature level fusion step to reduce the dimension of feature vectors for improving the recognition performance. Our proposed method is validated by forming three datasets using ORL, BANCA, FERET face databases and CASIA, UBIRIS iris databases. The results based on recognition performance and ROC analysis demonstrate that the proposed matching score level fusion scheme using Weighted Sum rule, tanh normalization, iris-FVF and facial features extracted by LBP achieves a significant improvement over unimodal and multimodal methods. Support Vector Machine (SVM) and t-norm normalization are also used to improve the recognition performance of the proposed method.
TL;DR: A new fully automatic system is presented to find the location of FAZ in fundus fluorescein angiogram photographs based on digital curvelet transform (DCUT) and morphological operations.
Abstract: In order to achieve early detection of diabetic retinopathy (DR) for the sake of preventing from blindness, regular screening using retinal photography is necessary. Abnormalities of DR do not have uniform distribution over the retina. Certain types of abnormalities usually occur in specific areas on the retina. The distance between lesions, such as micro-aneurysms, and the foveal avascular zone (FAZ) is a useful feature for later analysis and grading of DR. In this paper, a new fully automatic system is presented to find the location of FAZ in fundus fluorescein angiogram photographs. The method is based on two procedures: digital curvelet transform (DCUT) and morphological operations. Firstly, end points of vessels are detected based on vessel segmentation using DCUT. By connecting these points in the selected region of interest, FAZ region is extracted. Secondly, vessels are subtracted from the retinal image, and morphological dilatation and erosion are applied on the resulted image. By choosing an appropriate threshold, FAZ region is detected. The final FAZ region is extracted by performing logical AND between two segmented FAZ. Our experiments show that the system achieves, respectively, the specificity and sensitivity of (>98 and >96 %) for normal stage, for mild/moderate non-proliferative DR (NPDR) (>98, and >95 %) and for Sever NPDR + PDR (>97 and >93 %).
TL;DR: This paper studies the stochastic behavior of the least mean fourth (LMF) algorithm for a system identification framework when the input signal is a non-stationary white Gaussian process and a theory is developed based upon the instantaneous average power and the instantaneousaverage squared power in the adaptive filter taps.
Abstract: This paper studies the stochastic behavior of the least mean fourth (LMF) algorithm for a system identification framework when the input signal is a non-stationary white Gaussian process The unknown system is modeled by the standard random-walk model A theory is developed which is based upon the instantaneous average power and the instantaneous average squared power in the adaptive filter taps A recursion is derived for the instantaneous mean square deviation of the LMF algorithm This recursion yields interesting results about the transient and steady-state behaviors of the algorithm with time-varying input power The theory is supported by Monte Carlo simulations for sinusoidal input power variations
TL;DR: Some directional statistical features—including the mean and standard deviation of the local absolute difference—are integrated into the feature extraction to improve the classification ability of the extracted features.
Abstract: Local Binary Pattern (LBP) has achieved great success in texture classification due to its accuracy and efficiency. Traditional LBP method encodes local features by binarying the difference in local neighborhood and then represents a given image using the histogram of the binary patterns. However, it ignores the directional statistical information. In this paper, some directional statistical features—including the mean and standard deviation of the local absolute difference—are integrated into the feature extraction to improve the classification ability of the extracted features. In order to reduce estimation errors of the local absolute difference, we further utilize the least square estimate technique to optimize the weight and minimize the local absolute difference, which leads to more stable directional features. In addition, a novel rotation invariant texture classification approach is presented. Experimental results on several texture and face datasets show that the proposed approach significantly improves the classification accuracy of the traditional LBP.
TL;DR: A novel blind unmixing algorithm, constrained kernel nonnegative matrix factorization, which obtains the endmembers and corresponding abundances under nonlinear mixing assumptions, and two auxiliary constraints are introduced into the algorithm in order to improve its performance.
Abstract: Spectral unmixing has been a useful technique for hyperspectral data exploration since the earliest days of imaging spectroscopy. As nonlinear mixing phenomena are often observed in hyperspectral imagery, linear unmixing methods are often unable to unmix the nonlinear mixtures appropriately. In this paper, we propose a novel blind unmixing algorithm, constrained kernel nonnegative matrix factorization, which obtains the endmembers and corresponding abundances under nonlinear mixing assumptions. The proposed method exploits the nonlinear structure of the original data through kernel-induced nonlinear mappings and one need not know the nonlinear model. In order to improve its performance further, two auxiliary constraints, namely simplex volume constraint and abundance smoothness constraint, are also introduced into the algorithm. Experiments based on synthetic datasets and real hyperspectral images were performed to evaluate the validity of the proposed method.