Earticle

With the development of the theory of wavelet transform, biorthogonal wavelet filter banks with linear phase and compact support characteristics are widely used in signal and image processing. We study the multiscale edge detection using anti-symmetrical biorthogonal wavelet filter banks with the same even length based on the analysis of the properties of this wavelet filter banks. The steps of multiscale edge detection based on anti-symmetrical biorthogonal wavelet are introduced in detail. Experimental results on test images show that compared with the anti-symmetrical biorthogonal wavelet, the proposed anti-symmetrical biorthogonal wavelet filter banks with the same even length have better performance in terms of image edge detection. Therefore, the anti-symmetrical biorthogonal wavelet filter banks with the same even length are more suitable for image edge detection.

In this paper, a simple active image authentication scheme for the compressed images of block truncation coding (BTC) is proposed. In this scheme, the authentication codes of the compressed blocks are generated from the random value induced by the random seed. The authentication code of each compressed block is embedded into the bit map. The bit length of each authentication code can be chosen according to the user’s requirement. The experimental results indicate that the proposed scheme detects the tampered areas clearly and keeps good image qualities of the embedded images. Meanwhile, a low computational cost is required in the proposed scheme.

Scanned topographic maps of cities may have uneven illumination and creases. They need to go through color dodging. Due to the unsatisfying result by homomorphic filter of Retinex theory and the low efficiency when color dodging, this paper proposes an improved color dodging for scanned topographic maps based on Retinal-cortex theory. The main idea is to introduce Fastest Fourier Transform In The West (FFTW) to get the illumination component of the image on the basis of homomorphic filter of the Retinal-cortex algorithm, and use homomorphic filter to get the reflectance component and get the final image by linear stretching treatment. Experiments have proved that this method keeps the details of the image while erasing the creases so that the brightness distribution is improved.

Due to the significant increase of line-to-ground capacitive current, for urban distribution networks mixed by overhead lines and a larger number of cables, the higher electromagnetic transient overvoltage presents not only always on fault phase and neutral point, but also on non-fault phases when some intermittent arc grounding faults occur. Some arc-suppression coil which has a relatively smaller overcompensation tuning-off degree is mainly used to compensate the line-to-ground capacitive current, reduce the arc residual current and reignition times, and then increase the self-restoration probability after a single phase intermittent arc grounding fault occurs. At the same time, according to the characteristics and performance requirements of distribution systems, a relatively smaller shunt resistance is mainly used to quickly release the charge of line-to-ground capacitance, improve the decay rate of overvoltage of fault phase, non-fault phases and neutral point respectively, reduce the rising speed of neutral point recovery voltage and its amplitude after the intermittent arc dies out, so that the transient overvoltage can be restrained effectively. Results of transient numerical analysis and ATP simulation have demonstrated the mode of neutral grounding via arc-suppression coil with shunt resistance is an optimum choice which is applicable to 35 kV urban distribution networks mixed by overhead lines and cables.

A new steganographic algorithm is presented using pixel mapping method based on eight queens and number of ones in pixel intensity value. In our method, we are sequentially selecting 81 pixel blocks to embed message bit. The method works with randomizing the bit selection in the 81 pixel block using eight queens’ solutions. This approach finds relationship between secret message and cover image according to the pixel mapping table. In this paper, we have presented a secure image steganography technique with acceptable levels of imperceptibility and distortion in the cover image. In presenting algorithm, security is considered by the randomize selection of eight queen’s solutions based on a seed value. The presented algorithm not only provides high level security, but also produces high capacity and good imperceptibility. The algorithm has been tested with different image file formats like BMP, TIFF and PNG.

The study on classification methods of hyperspectral image is a focal growing area in remote sensing applications because the wide spectral range, providing a very high spectral resolution, allows the detection and classification surfaces and chemical elements of the observed image. Semi-supervised learning method which takes a large number of unlabeled samples and minority labeled samples, improving classification and predicting the accuracy effectively have been a new research direction. In this paper we proposed a new semi-supervised classification method of hyperspectral image based on combining Renyi entropy and multinomial logistic regression algorithm. The multinomial logistic regression was performed to describe a direct relationship between the selected sample as and their category. A lot of unlabeled samples are constantly added to the sample data using Renyi entropy algorithm. The test analysis of image classification in test area showed the advantages of classification method based on combining Renyi entropy and multinomial logistic regression algorithm for hyperspectral remote sensing image.

The aim of image enhancement is to or to provide ‘better’ input for other improve the interpretability or perception of information in images for human viewing automated image processing techniques. Various Histogram Equalization techniques like CHE, GHE, BBHE, DSIHE, RMSHE and Multi-HE techniques are used for processing the image input to enhance its output. This paper provides a review over the modification of the brightness preserving dynamic histogram equalization technique to improve its brightness preserving and contrast enhancement abilities while reducing its computational complexity. There are many modified technique related to brightness preserving dynamic Histogram Equalization that uses statistics of digital images for their representation and processing are discussed here. Representation and processing of images in the spatial domain enables the technique to handle the inexactness of gray level values in a better way, resulting in improved performance. This algorithm enhances image contrast as well as preserves the brightness very effectively. Some images are not available to good quality, so these algorithms are used for image enhancement to improve the quality of the image.

In this paper, we propose a fast and robust tracking method based on reversed sparse representation. Be different from other sparse representation based visual tracking methods, the target template is sparsely represented by the candidate particles which are gotten by particle filter. In order to improve the robustness of the method, we use a target template set. Meanwhile, a two level competition mechanism is also introduced. In the first level, each target template is sparsely represented and all the candidate particles compete with each other by a similarity calculation, which is based on sparse coefficients. Then, the winners construct a target candidate set. In the second level, all the target candidates in the target candidate set compete with each other and the one which is the most similar to the template set is considered as the target. In addition, a template set update strategy is proposed to adapt the appearance variations of the target. Experimental results on challenging benchmark video sequences demonstrate that the proposed tracking algorithm performs favorably against several state-of-the-art methods.

This paper focuses on the problem that video-GIS has a huge amount of data, which leads to high transmission resource consumption, and introduced attention calculation model to video GIS coding to optimize the coding method so the compression efficiency will be improved as well. Specifically, on the one hand, it will use optical flow technique to calculate the specific location and motion vector of the movement point set of each frame, and reduce the amount of the movement point set integrate with the features of video GIS, then calculate the mask matrix of the foreground movement, and finally compute the Discrete Cosine Transform (DCT) compression integrate with the mask matrix to realize the optimization of the frame macro block level coding. On the other hand, it will calculate the attention of the video frame using the specific location and the motion vector of the movement point set as the primary indicator, and optimize the coding of the video of frame-level according to the calculation result. By experimental verification, the efficiency of video GIS compression is enhanced by the introduction of the features of the video GIS and the cognitive attention theory

This paper proposes image compression in different color spaces using hybrid wavelet transform. To generate hybrid wavelet transform Discrete Kekre transform (DKT) and Discrete Cosine transform (DCT) are selected as component transforms. Due to high energy compaction property, DCT is selected as local component transform that contributes to local features of an image. Hybrid wavelet transform extracts features of both the component transforms and hence gives less error and better image quality. Component transforms of different sizes are selected to generate hybrid wavelet of size 256x256 and applied on images. In RGB color space 16-16 combination i.e. hybrid wavelet generated using DKT 16x16 and DCT 16x16 gives least error than other combinations like 8-32, 32-8 and 64-4. RMSE, MAE, AFCPV and Structural Similarity Index (SSIM) are the error metrics used to measure reconstructed image quality. Different color spaces have been used to observe the performance of this hybrid wavelet transform. In KLUV color space minimum RMSE and MAE is observed than RGB, YUV, YCbCr, XYZ and YIQ color space. Whereas RGB color space gives lowest AFCPV than other color spaces using 16-16 component size. Hence SSIM is used to eliminate this inconsistency in these traditional error metrics. KLUV color space gives highest SSIM 0.998 which is closest to maximum one proving it as a better choice than other color spaces.

We address the issue of continuous tracking of the target in an environment covered by multiple cameras. In such a scenario, target handoff is a key problem. In this paper, we propose a novel target handoff method based on appearance model inheriting and learning. The appearance model is initially learned by sparse representation using the tracking results in the first camera. The next camera inherits the appearance model for target handoff and updates it after getting the whole tracking results. Then, the appearance model is transferred to the following camera. By the appearance model inheriting and learning, the appearance model can describe the target more and more precisely, which will make the target handoff more accurately and effectively. We also demonstrate the performance of our method on several video surveillance sequences.

Semi-supervised constraint scores, which utilize both pairwise constraints and the local property of the unlabeled data to select features, achieve comparable performance to the supervised feature selection methods. The local property is characterized without considering the pairwise constraints and these two conditions are introduced independently. However, the pairwise constraints and the local property may contain conflicting information. In this paper, we utilize the conflicting information to improve the local property. Instead of characterizing the local property by all neighbors, samples which do not appear in the cannot-link constraints can be used. A performance indicator, called neighborhood-cannot-link (NC) coefficient, is proposed to measure the improvement of the local property. We use the improved local property and the pairwise constraints to perform semi-supervised constraint scores algorithm. Experiments on several real world data sets demonstrate the effectiveness of the methods.

In stereo vision, two images of a 3D scene are acquired from two viewpoints. One of the objectives of stereo vision work is to recover the 3D structure of the scene. Epipolar geometry describes the relationship between the images, and the essential and fundamental matrices are the algebraic representations of this geometry. The most important feature of these matrices that is emphasized in the literature is that they are independent of the scene structure. This article illustrates—empirically and theoretically—that the fundamental matrix depends on the scene structure and demonstrates that the matrix in 0lrFmm not only represents a relationship between corresponding points of the two views but also represents a relationship between other non-corresponding points. Furthermore, we show empirically that the equation 0lrFmm does not hold for any pair of corresponding points. In scenes with objects of different depths, the value of lrFmm depends on the depths of the 3D points and increases proportionally with an increasing baseline.

In the surface measurement system, the data acquisition is key part and the anti-aliasing analog filter is necessary. This paper deals with the design of the anti-aliasing analog filter. Based on the principle of anti-aliasing filtering, the parameters of filter are computed, the filter circuit is designed, and the frequency characteristic is drawn. Combined with digital filter, the filtering can maintain the low frequency components very well and suppress effectively the high frequency signals in the original surface profile, which reduces the distortion caused by noise and makes the filtering effect better. This method has been applied in the surface measurement system and the actually measured data verified the performance of the filter.

Orthogonal Frequency Division Multiplexing (OFDM) is an efficient method of data transmission for high speed communication systems. One of the challenging issues for OFDM system is its high Peak-to-Average Power Ratio (PAPR) which causes inefficient use of the High Power Amplifier (HPA) and could limit transmission efficiency. In this paper we analyze the effects of PAPR on HPA and its reduction technique based on GCD matrix In Selected Mapping (SLM) method. Simulation results shows that the new phase sequence based SLM method has good performance in PAPR reduction.

Voice activity detection (VAD) is the key of voice recognition, voice synthesis and speech-sound enhancement.For the sake of improve the accuracy and robustness of speech endpoint detection system. Combining the advantages of adaptive genetic algorithm (AGA) and improved radial basis function network (RBF) defects in existing learning methods. This paper presents a comprehensive detection method-- Adaptive genetic algorithm radial basis function network. This method uses adaptive genetic algorithm to simultaneously optimize the center, the width and the structure of RBF network. The method using wavelet analysis to extract the characteristics of the speech signal, use them as an input amount to the radial basis function networks. Establish voice detection system model, this method enhance the accuracy of the detection system and has better robustness.

Brain image segmentation is one of the most important parts of clinical diagnostic tools. However, accurate segmentation of brain images is a very difficult task due to the noise, inhomogeneity and sometimes deviation in brain images. Wells model incorporates the brain image segmentation and inhomogeneity correction into one framework to overcome influences from the intensity inhomogeneity and obtain good segmentation performance. However, the classical Wells model did not take spatial information into account, so its segmentation results are sensitive to the noise. In order to overcome this limitation, the MRF theory and the nonlocal information are used to construct a nonlocal Markov Random Field. With this nonlocal MRF, the improved Wells method can obtain much better segmentation results. The experimental results show that our method is robust to the noise and is able to simultaneously keep the image edge and slender topological structure very well.

The exposure fusion technique is very useful for directly fusing multi-exposure images into a high quality image, without the physically based high dynamic range (HDR) production step. In this paper, we present an improved method for multi-exposure images fusion using multi-resolution Laplacian pyramid weighted blending. We propose a Weight Modification Factor (WMF) to modify the original weight map for each image guided by contrast, saturation and well-exposedness of each pixel in the image, and to enhance the weight of the informative pixels in ultra-bright or ultra-dark areas. Compared with previous methods, our approach preserves more useful details for scenes with very large dynamic range and achieves no artifacts. Several objective quality metrics prove the advantages of our method.

Disparity estimation for still images has attracted much interest and acquired many promising results. However, simply applying these methods to produce a disparity sequence may suffer from the undesirable flickering artifacts. These errors not only distinctly decrease the visible quality of the synthesized video, but also significantly reduce the coding efficiency of the disparity sequence. In this paper, a novel temporal consistency enhancement algorithm based on Guided Filter and Temporal Gradient (GFTG) is proposed. The flickering artifacts and noises are effectively removed and the edges of objects are well preserved. Both quantitative and qualitative evaluations show that the spatio-temporal consistency has been highly improved by utilizing our approach.

Various kinds of images and pictures are required as sources of information for analysis and interpretation. When an image is converted from one form to another such as scanning, transmitting, digitizing, storing etc., degradation occurs to the output image. Hence, the output image needs to be enhanced in order to be better analyzed. Denoising is the one of the pre processing technique in digital image processing. This paper investigates the performance of four denoising methods for removing the High Density Impulse Noise. They are Adaptive Bilateral Filter (ABF), Fuzzy Peer Group Filter (FPGF), Switching Bilateral Filter (SBF), and Boundary Discriminative Noise Detection Filter (BDND).The performance of the above four filters is compared by using five performance metrics. They are Peak-Signal-to-Noise-Ratio, Mean Square Error and Root Mean Square Error. The Experimental results show that the BDND filter based denoising method performs well than the other three methods.

In order to use pulse coupled neural networks (PCNN) for precise automatic image segmentation, we propose an improved PCNN model. We first establish a connection weight matrix based on the image local gray correlation and on the Euclid distance. We then used the minimum variance ratio criterion to automatically determine PCNN cycle times, and achieve automatic image segmentation. The simulation results show that this method can automatically determine the number of iterations PCNN, and that it is highly feasible and better segmentation effect.

The use of fingerprint for human identity management has been on the rise lately. Reasons adduced for this include its high level of uniqueness, availability, consistency and universality. The task of human identity management based on fingerprint technology involves a number of processes which include enrolment, enhancement, feature and singular points detection and extraction and pattern matching. Detection and extraction of genuine and reliable feature and singular points are paramount for reliable pattern matching. The limitations of some existing fingerprint singular point detection algorithms include inaccurate detection and failure with some fingerprint pattern and poor quality images. In this paper, a modified Poincare Index fingerprint singular point detection algorithm is proposed for resolving these limitations. The results of the experimental study of the new algorithm on Dataset DB1 of FVC2000 standard fingerprint database show that the new algorithm reliably and adequately detected singular points from fingerprints of all patterns and qualities.

In this paper, we firstly propose a new image restoration model including non-smooth l1-norm data-fidelity term and non-smooth l1-norm regularization term based on the bilateral total variation regularization. Secondly, we prove the existence of minimal solutions of our proposed energy functional model. Thirdly, we consider the convergence of the discrete numerical algorithm, and obtain that the limit point of the solution sequence is the minimal point of our proposed energy functional. Finally, we give some experimental simulation results in the case of the single noisy image without blurring, multiple different noisy images without blurring, single noisy image with blurring, and multiple different noisy images with different blurring, respectively. The restoration results show our model works effectively.

Object Recognition System of Sonar Image plays an important role in the field of underwater defense According to pattern recognition theory, principle processes of a typical object recognition system is introduced . To achieve robust performance, image de-noising, sharpening, binary transformation, edge detection and other image processing techniques are discussed . The paper presents a novel object recognition system using multiple invariant moments as the main feature of the object, and the detected feature is trained by BP neural network so that the classification error can be minizied. Finally, we implemented the proposed approach by visual air plane recognition, the experimental results demonstrates the robustness and efficiency of the presented approach.

To improve the social life of drug resistant epilepsy persons, a patient specific algorithm is needed that can predict seizures based on EEG with high sensitivity and specificity before the occurrence of a seizure. This algorithm predicting the seizure occurrence from Inter-ictal (seizure free) and pre-ictal (before seizure) transition. In this algorithm features are extracted by Fourier Bessel Expansion from inter-ictal and pre-ictal EEG signals. A neural network using back propagation algorithm is implemented for classification of epileptic states. The performance of algorithm is evaluated based on three measures, sensitivity, and specificity and classification accuracy. The results illustrate that the algorithm can predict seizures of two subjects before five minutes with an accuracy of 99.6%

An online surface inspection system m of hot rolled strips is introduced. This system is designed t o detect such main Surface defects on hot rolled strips as scar, scratches, pits, water drops Cracks. Cross hatchings, and so on. Multiple CCD area scan cameras are adopted to capture images of strip surface simultaneously, and all the images are processed by parallel computation system Real-time, which is supported by fast image process techniques and parallel computation techniques, in order to snap main defect regions on the surface of strips. At last, the defects will be classified to several types. The application of the system to practical production line shows that it can detect main defects of hot rolled strips more effectively than traditional method, and it is easily to be maintained.

IMM (Interacting Multiple Model) algorithm is widely used in target tracking, and its basic principle is described in detail at first. However, the IMM algorithm fails to obtain the prior probability of model conversion quickly and accurately when tracking for target. In this paper, an improved IMM algorithm based on ANFIS (the adaptive neural fuzzy inference system) is proposed. The improved algorithm can update the value of system noise covariance in real-time by ANFIS module through observing the coefficient of system noise covariance. Consequently, the probability of model conversion can be obtained more quickly and accurately. Then, the comparison and analysis of the experiment results between the original IMM algorithm and the improved one have been carried out. The experiment results show that the reaction rate for maneuvering target tracking is significantly boosted and tracking error is obviously reduced because the improved algorithm can update the value of system noise covariance in real-time and improve the system adaptability.

Affective state detection, as an emerging field of artificial intelligence, is the key to designing effective natural human-computer interaction, especially for e-learning. It will be helpful to make the computer understand learners’ perceptions and provide appropriate guidance, just like teachers in traditional face-to-face classroom learning. Puzzlement is the most frequent non-neutral affective state in learning, and it is usually a sign that learners need more information and guidance. In this paper, we explore a machine learning approach for puzzlement detection from natural facial expression. We use active appearance models (AAMs) to decouple shape and appearance parameters from the face video sequences. Support vector machines (SVMs) are utilized to classify puzzlement and non-puzzlement with several features derived from AAMs. Using a 10-fold cross validation, we achieve the highest recognition rate of 98.9%. Experimental results indicate the feasibility of automatic frame-level puzzlement detection.