Earticle

All phase digital filter (APDF) is a new type of linear phase FIR digital filter that was proposed in recent years. Firstly, the theory of biorthogonal wavelet transform and windowed APDF based on DCT is expounded and the relationship between them is discussed. Secondly, a novel algorithm is proposed to implement biorthogonal wavelet transform by using windowed APDF based on DCT. As an important application of biorthogonal wavelet transform, multi-resolution analysis of 2-D image signal could be used to test the feasibility and applicability of the proposed algorithm. Finally, the simulation results using MATLAB tool are shown in this paper and the analysis indicates that the proposed method performs well.

Iterative gradient descent algorithms to solve algebraic linear equations are well known in the literature. These algorithms can be interpreted as dynamical closed loop control systems, where the step sizes are the control variables that can be optimally calculated by the Liapunov Control Functions (CLF) and Liapunov Optimizing Control (LOC) methods. In this paper matrix versions of gradient descent algorithms are deduced, including an unpublished matrix form of the Barzilai-Borwein algorithm. The step sizes (control variables) are also calculated by CLF/LOC. The main utility of these matrix algorithms is in image deblurring in the cases where the matrix that represents the blurring process is too large to be stored in the memory of the majority of conventional computational systems.

Due to the fact that theoretical analysis and instrument construction for antenna array of synthetic aperture imaging radiometer (SAIR) are both complicated, and designers usually hope to predict the imaging effect and analysis the influence of relevant parameters of SAIR before the system design. The imaging simulation is a very useful method in the system design of SAIR. This paper is devoted to establishing an accurate imaging model for simulating the process of near-field millimeter wave SAIR. In this model, the target radiation signals and received signals of receivers are represented by the accurate signals in the time domain, which improves the efficiency of this simulation model significantly. The visibility function is collected by the cross-correlation between I/Q signals of the antenna pairs just as the imaging process of the practical SAIR. Some characteristics (such as the coherence between targets, the resolution and no-aliasing FOV of SAIR and so on) are verified by the corresponding 1D and 2D simulation experiments, and the effectiveness of this imaging model is also tested by these simulation experiments. The simulation experiment results show that this model is an efficient, accurate imaging simulation model, and can be employed in the system design of near-field millimeter wave SAIR.

Content-based image retrieval is a very important area of research nowadays. Content Based mage Retrieval (CBIR) is a technique which uses visual features of image such as color, shape, texture, etc. CBIR technologies provide a method to find images in large databases by using unique descriptors from a trained image. A lots of research works had been completed in the past decade to design efficient image retrieval techniques from the image or multimedia databases. Large number of retrieval techniques has been introduced, but there is no universally accepted feature extraction and retrieval technique available. In this paper, we present a study of various content-based image retrieval systems and their behaviour, texture analysis and various feature extraction with representation.

Two novel methods for image feature extraction based on fractional differentiation are presented in this paper. The first method is the feature extraction of fusing multi-direction CRONE operators. In this method, the fractional differential CRONE mask is generalized to eight directions at first for extracting image features; then the extracted features are tested by the statistic method and fused by the gradient ratio, so that the outlines of the objects in the image are obtained. In order to extract the detail feature information in the image effectively, the second method, the ‘S+Z’ extraction combined with the space-filling curves, is presented. By introducing the space-filling curves, the ‘S’ curve and the ‘Z’ curve, and making full use of the neighborhood information of image pixels, the detailed features of the objects in the image are obtained. The experiment results show that our methods can obtain satisfactory image features.

The traditional Snake model and GVF-Snake model set high requirements on noise and initial contour in wood cell contour extraction. To solve this problem, on the premise of considering the image texture and gray-scale information, the area information is directly introduced into the active contour extraction model through force equilibrium equation. Experiments show that the contour extracted with this method is not only more close to real cell contour, but also improved in anti-noise property. In particular, in the convergence of high noise and deep sunken areas, it has some advantages not found in other traditional methods.

The survey paper explains about the extraction and retrieval of personal name alias using various techniques from the web with the help of web crawls. The existing methods help to improve the depth of knowledge relevant to alias extraction and retrieval process. It also describes about how the aliases are ranked, then page counts on the web, word co-occurrence using anchor text and techniques like term frequency (tf), inverse document frequency (idf), log likelihood ratio. Chi-squared tests etc.., are used for measuring the association and similarities between words. The existing method consists of pattern extraction algorithm or string matching algorithm for extracting patterns from snippets instead of using these algorithms. The survey helps to discover a proposed method as graph mining to extract personal name aliases from the web.

Granulation of information appears in many areas, such as machine learning, evidence theory, and data mining. Granular computing is the core research field in granulation of information. It is an effective tool for complex problem, massive data mining and fuzzy information processing. In the basis of principle of granularity, we aim to study the granular decomposing method in granules space based on rough set. Moreover, the criteria conditions for attribution necessity and attribute reduction are proposed. Finally, the corresponding equivalence is proved to traditional rough set theory. It will lay the foundation for attribute reduction under the granular representation in rough set.

DNA microarray technique can detect tens of thousands of genes activity in cells and has been widely used in clinical diagnosis. However, microarray data has the characteristics of high dimension and small samples, moreover many irrelevant and redundant genes also decrease performance of classification algorithm. Feature gene selection is an effective method to solve this problem. This paper proposes a hybrid feature gene selection method. Firstly, a lot of irrelevant genes from original data were eliminated by using reliefF algorithm, and the candidate feature genes subset is obtained; Secondly, Fuzzy neighborhood rough set with information entropy which deals directly with continuous data is proposed to reduce redundant genes among genes subset above. Here, differential evolution algorithm is used to optimize radius before reduction by using fuzzy neighborhood rough set, because radius of neighborhood greatly affects reduction performance. The simulation results on six microarray datasets indicate that our method can obtain higher classification accuracy by using as few genes as possible, especially feature genes selected are important for understanding microarray data and identifying the pathogenic genes. The results demonstrated that this method is effective and efficient for feature genes selection.

Multimedia security is extremely significant concern for the internet technology because of the ease of the duplication, distribution and manipulation of the multimedia data. The digital watermarking is a field of information hiding which hide the crucial information in the original data for protection illegal duplication and distribution of multimedia data. This paper presents a survey on the existing digital image watermarking techniques. The results of various digital image watermarking techniques have been compared on the basis of outputs. In the digital watermarking the secret information are implanted into the original data for protecting the ownership rights of the multimedia data. The image watermarking techniques may divide on the basis of domain like spatial domain or transform domain or on the basis of wavelets. The spatial domain techniques directly work on the pixels and the frequency domain works on the transform coefficients of the image. This survey elaborates the most important methods of spatial domain and transform domain and focuses the merits and demerits of these techniques.

Tracking-Learning-Detection (TLD), also known as Predator, has become one of the most popular state-of-the-art algorithms in the domain of visual tracking in recent years. It has demonstrated outstanding performance in the application of long-term tracking of a human face in unconstrained videos. In this paper, we address the problem of tracking a single ship in inland waterway CCTV videos given its location in the first frame and no other prior information. Firstly, we deeply analyze Predator in a perspective of control system and point out the search strategy in detection procedure is the most time-consuming part in Predator system. Secondly, in order to speed up the whole pipeline, we propose a novel motion model based on extended particle filter with orthogonal design. Due to the power of particle optimization and re-combination with orthogonal design, we can relate the motion of object of interest better and obtain the most likely candidates of object regions more effectively. Finally, both qualitative and quantitative evaluations on numerous challenging CCTV videos demonstrate that the proposed algorithm achieves favorable performance in terms of efficiency and accuracy.

In surveillance systems, the storage requirements for video archival are a major concern because of recording of videos continuously for long periods of time, resulting in large amounts of data. Therefore, it is essential to apply efficient compression techniques for compressing surveillance video. The techniques used for the general video compression may not be the efficient technique for the compression of surveillance video because of the use of static camera as compared to moving camera in general purpose videos. Generally surveillance video consist of multiple objects, smaller in size as compared to the background and they have frequents occlusion with each other. In this paper a new object based motion estimation and compensation technique for surveillance video compression is proposed. Background differencing and summing technique (BDST) is used for the segmentation of the moving objects. This technique not only identifies moving object but also the maximum distance moved by the object in given group of frames. A bonding box is created based on the movement of the object in order to segment the moving objects. For exploiting the temporal redundancy, the motion estimation and compensation is carried out for the bonding box region only. The multiresolution property of discrete wavelet transform is used for the motion estimation and compensation. Experimental results show that the approach achieves high compression ratios compared to MPEG-2 compression.

The concept of the quaternion is useful for colour image processing and recognition. This paper introduces quaternion discrete Tchebichef moments (QTM), which use the traditional Tchebichef moments (TM) of each colour channel to describe colour images. A set of invariants that are invariant to translation and scale transformations is introduced for colour object recognition and image classification. A theoretical framework is provided for the recognition of colour face images by combining the proposed quaternion Tchebichef moment functions with the sparse representation classification (SRC) strategy for improving recognition despite partial occlusions. Simulation results on standard colour face databases demonstrate the effectiveness of the proposed algorithm, even when the images include Gaussian or pepper-and-salt noise.

In this paper a watermarking technique using hybrid wavelet transforms obtained from sinusoidal and non-sinusoidal component orthogonal transforms is proposed. Sinusoidal transform DCT and non-sinusoidal transforms Walsh, Haar and Discrete Kekre Transform are used to generate hybrid wavelet transforms namely DCT-Walsh, Walsh-DCT, DCT-Haar, Haar-DCT, DCT-DKT and DKT-DCT. Size of each component transform matrix is varied suitably from 4, 8, 16, 32, and 64 to generate hybrid wavelet transform matrix for host and watermark. The best size combination is further applied column wise and row wise to host and watermark and to embed the watermark middle frequency regions of host is selected. Embedding is first done without sorting the hybrid wavelet transform coefficients of host and watermark and then sorting is applied to observe the difference in the achieved robustness. Performance of proposed technique is evaluated against various attacks to decide whether sinusoidal transform when used as base transform matrix or local transform matrix is more robust.

This paper aims to present a self-adaptive global particle swarm optimization (SGPSO) algorithm for solving unconstrained optimization problems. In the new algorithm, the inertia weights are generated based on Gaussian distribution, which is helpful to improve the diversity of the population. In addition, the worst particle is updated by averaging the other particles, which is beneficial to improving the quality of the population. Finally, a global disturbance is adopted to increase the convergence rate of SGPSO. In the disturbance process, a disturbance factor is utilized to control the searching ranges of the population, which can effectively keep a balance between the global exploration and local exploitation. Twenty well-known benchmark functions are considered to evaluate the performance of SGPSO, and 50 runs are implemented in each case. Numerical experiments and comparisons demonstrate that SGPSO is superior to the other three algorithms according to means, standard deviations and convergence rate.

Low visibility is the primary cause of the driving accidents. Current image enhancement methods can not effectively improve driving visibility. An image enhancement method based on symmetrical logarithm transformation is proposed to improve this situation. Symmetrical logarithm transformation based on two-scale luminance distribution features is used to obtain an enhanced image. The advantages of proposed method include its self-adaptive ability to varied brightness, low computational complexity and parameter-free input. Experimental results show that the performance of proposed method is more robust and effective than that of other state-of-the-art image enhancement methods.

Off-line recognition of text plays a significant role in several applications such as the automatic sorting of postal mail or editing old documents. The recognition of Arabic handwriting characters is a difficult task owing to the similar appearance of some different characters. Most researchers have presented methods that recognise isolated characters. However, recognition of all shapes of Arabic handwritten characters still remains a great challenge. The selection of the methods for feature extraction and classification remain the most important step in achieving high recognition accuracy. The purpose of this paper is to compare the effectiveness of DCT and DWT in capturing discriminative features of all shapes of Arabic handwritten characters including overlapping characters with ANN and HMM in the classification stage. Since, the recognition of handwritten characters is an important step in the recognition of a word after segmentation, this paper ascertains the effectiveness of these techniques in capturing useful information and, hence, achieving more accurate recognition results. This work has been tested with HACDB database containing 6,600 shapes of Arabic characters. The results have demonstrated that the feature extraction by DCT with ANN yields a higher recognition rate.

In this paper, we propose a novel approach to explore image denoising for patch based image process. The importance measurement model of Rough Entropy and the importance reduction method of wavelet coefficients are given. This paper combines the rough set theory with the denoising method of wavelet threshold, regarding the high-frequency information blocks in the transform domain as similar ones, and adopting importance Reduction Methods to contract the coefficients. The simulation results show that this method is effective.

Biometrics has been one the main security solutions in almost every type of infrastructure (whether critical or non-critical) ranging from the main doors at home, libraries to the critical infrastructures like banks and airports. Despite the forceful impetus on research on biometric security that has taken biometrics from one simple level to much higher levels of security, there are still some open challenges in this field of security that need to be filled. Among all those challenges and loopholes, the security of template is of the most important concern. The reason for this is that we don’t want any identity compromises. If a biometric template in the database of the system of a person is compromised that consequently would mean identity theft of that person. This paper proposes a novel method that uses two different biometric data from the same person for making a biometric template against each person. The two biometric modalities that have been used in our work are left and right iris using best features. The features and verification of the proposed system has been done using MATLAB.

Medical diagnosis based on vein characteristics attracted large attentions. The change and some feature of vein will reflect the diseases which can be observed from the diagnosis from specialists. Thus, it is very important to process the vein image for medical diagnosis. In order to process the vein image efficiently and effectively, this paper introduces a vein image processing approach based on the morphology principle, which is suitable for examining the shape of veins. The vein images captured from human fingers are used for the experiments. It is observed that the vein shape could be examined clearly so that any diseases or potential risks could be diagnosed by the doctors, comparing with traditional image processing approaches.

Nowadays, with the rapid growth in information technology more and more images and data are available on the internet. So there is a need to provide some kind of authentication to such important data. When the sender transmits the image to the receiver, there may be intruders present in between who may capture the image. After capturing the image the mage the intruder may view the meaningful content in the image. This may not be the problem in some cases. But if we consider security applications like medical and military images then such distortion is unacceptable. To avoid misuse or loss of information several reversible data hiding techniques (RDH) are implemented. This paper gives details on watermarking, LSB, Histogram and RDH using optimum Thresholding with related results.

This paper presents an accurate large displacement optical flow estimation approach by adaptively integrate the local feature match. Despite coarse-to-fine warping approach can handle large displacement optical flow; however, there is inherent limit for small object with large motion. And recently integration of feature match to the variational framework has relaxed the limit, but raised another problem of ambiguous feature matching due to poor feature descriptor. Address the aforementioned problem, in this paper we propose an adaptive integration approach of local feature match. The essence is that we only keep the robust feature and remove those unstable features (e.g, textureless region) to improve the flow accuracy. The adaptive approach substantially decreases the computational cost by remove uncertain features and leads to more robust performance by excluding unreliable matches. We qualitatively and quantitatively compared to the conventional flow methods on Middlebury and Sintel benchmark and show that we achieve more accurate and promising results.

In standard group search optimizer (GSO) algorithm, scroungers will converge to the similar position if the producer cannot find a better position than the old one in a number of successive iterations and the group may suffer from the premature convergence. In this paper, a hybrid GSO with differential evolution (DE) operator named DEGSO is proposed to enhance the diversity of standard group search optimizer. In this method, the standard GSO algorithm and the DE operator alternate at the odd iterations and at the even iterations. The results of the experiments indicate that DEGSO is competitive to some other evolutionary computation (EA) algorithms.

Today is an era of digital imaging. This can be viewed either in the field of photography or in the field of medical imaging. Digital imaging has improved the performance of picture quality. Detailed information can be recovered very quickly from any part of an image and this feature has become very useful in every field of imaging. This improvement in the field of medical imaging has given life to so many patients as diagnosis of disease has become very fast and easy. But many a times the image quality is not upto the mark, due to this reason; the doctors are not able to diagnose the disease. So this paper proposes a noval approach for improvement in quality of medical images using pixel reconstruction followed by Gabor filter enhancement technique. The experimental results are verified as improvement in PSNR of hexagonal pixel images as compared with square pixel images. The results show a large improvement in quality of digital imaging.

In recent years, more and more researchers' attention has been drawn to the sparse representation-based classification (SRC) method and its application in image analysis and pattern recognition, due to its good characteristics of high recognition rate, robustness to corruption and occlusion, and little dependence on the features selection etc. However, sufficient training samples are always required by the sparse representation method for the effective recognition. In practical applications, it is generally difficult to obtain sufficient training samples of the test targets, especially non-cooperative targets. So the key issues in the effective automatic target recognition (ATR) based on the sparse representation are to obtain sufficient training samples in different scales, angles, and different illumination conditions, and to construct an overcomplete dictionary with discriminative ability. In this paper, a novel sparse representation-based scheme is proposed for the automatic target recognition in the real environment, in which the training samples are drawn from the simulation models of real targets and the overcomplete dictionary is trained using structured sparse learning method. The experimental results show that the proposed method is effective for the automatic target recognition in the practical application, especially, where the desired features of the sparse representation method are kept.

Aiming at the impulse noise generated in capturing the images of insulator on power lines, a denoising method based on peer groups is proposed. The center pixel variance σcenter is defined, the minimum of neighborhood variance and center σmin is treated as threshold σmin and the peer group is determined by comparing the relation between absolute value of gray value difference and σmin . According to the size of peer group and its complement set, center pixel is estimated when noisy pixels exist in the neighborhood window. Otherwise, the size of window is adjusted adaptively and center pixel is estimated on the basis of mean value of non-noisy pixels within adjusted window. The experimental results show that the method can get a higher peak signal to noise ratio, IEF and SSIM when there is high density impulse noise in an image.

Millimeter wave radar echo signals often contain noise and clutter because of rain and fog’s influence on the performance of which, and its performance drop greatly. In recent years, bi-stable stochastic resonance and multi-scale wavelet decomposition theory received great attentions in the field of signal de-noising. This paper proposed a novel mechanism of stochastic resonance which is induced by multi-scale noise for weak signal detection in millimeter wave radar signal. Firstly by multi-scale wavelet decomposition, input signal which is in heavy background noise was decomposed to several signals with different frequencies. After that they were induced by contraction factors of each noise scale, and then were as the input signal of bi-stable system. Simulations of different parameters show that under suitable contraction factors, SNR of output signal can be improved greatly.

Manifold learning is an approach for nonlinear dimensionality reduction and has been a hot research topic in the field of computer science. A disadvantage of manifold learning methods is, however, that there are no explicit mappings from the high-dimensional feature space to the low-dimensional representation space. It restricts the application of manifold learning methods in many practical problems such as target detection and classification. Previously, some methods have been proposed to provide linear or nonlinear mappings for manifold learning methods. However, a disadvantage of all these methods is that the learned projective functions are combinations of all the original features, thus it is often difficult to interpret the results. Moreover, the dense projection matrices of these approaches lead to a high cost of computation and storage. In this paper, a sparse polynomial mapping approach is proposed for manifold learning. We first get the low-dimensional representations of the high-dimensional input data by using a manifold learning method, and then a 𝑙1-based simplified polynomial regression is used to get a sparse polynomial mapping between the high-dimensional data and their low-dimensional representations. In particular, we apply this to the method of Laplacian eigenmap and derive a sparse nonlinear manifold learning algorithm, which is named sparse locality preserving polynomial embedding. Experimental results on real-world data show the effectiveness of our approach.

Speech Signals have high range of variation in amplitudes and frequency. These acoustic signals with diverse properties are hard to recognize and filter if mixed with noise. To separate noise from original signal, the artifact peaks are separated from original signal and discarded. In this paper, the ICA method of signal denoising is used to differentiate the speech signal from periodic noise and Empirical Mode Decomposition method is proposed to generate the components of signal. The IMF(s) of signal is the non-linear descending order of frequency components that have been filtered for better SNR. Filtering with wiener filter has amended output but also results in loss of information. The selection of IMF(s) for signal regeneration when optimized using objective function of PSO, the information of original signal was dramatically preserved with suppressed noise. The system is tested on 4 example signals and proposed technique illustrates lower mean square error and higher SNR compared to wiener and ICA.