Narrow Results

The Bloomberg Carbon Clock is a web-based and real-time estimate of the global average atmospheric CO₂ level. The work was conceived to help non-scientists understand the seasonal nature of atmospheric composition, and offers a click-through explanation of the CO₂ data’s meaning, history, and implications. The Carbon Clock was designed to satisfy both scientific rigor and aesthetic appeal.

Edge-preserving image enhancement and noise removal are of great interest in medical imaging. This chapter describes schemes for noise suppression of magnetic resonance images using wavelet multiscale thresholding. To sufficiently exploit the wavelet interscale dependencies, we multiply the adjacent wavelet subbands of a Canny-edge-detector-like dyadic wavelet to form a multiscale product function where the significant features in images evolving with high magnitude across wavelet scales are amplified while noises are deteriorated, which facilitates an easy differentiation of edge structures from noise. Thereafter an adaptive threshold is calculated and imposed on the products, instead directly on the wavelet coefficients, to identify important features. Experiments show that the proposed scheme outperforms other wavelet-thresholding denoising methods in suppressing noise and preserving edges.

Local reconstruction, metal artifact reduction and image denoising is the main problems in Computerized Tomography. In this paper, we chiefly introduce the wavelet methods in CT to deal with these problems. Furthermore, we propose the integration of wavelet methods.

Wavelet transformation is in the last few years an emerging science that develops rapidly. As a mathematics tool, the wavelet technique has been made use of in numerous fields quickly. The Virtual Reality Training System of Vehicle Driving that we developed has used the wavelet transformation technique in portrait handling.

This research paper originally build up the Wavelet — Kalman Filtering Hybrid Estimating and Forecasting Algorithm (WKHEFA), which incorporates the advantages of both Kalman filtering and wavelet analysis. And it successfully forecasts the 30-minute trading volume of Shanghai Stock Exchange based on WKHEFA.

In this paper, a novel method of feature extraction in non-stationary lung sound signals is put forward. Four kinds of lung sounds data (collected in the state of normal, tracheitis, pneumonia and asthma respectively) were sampled from various subjects. We use db5 wavelet to do fast arbitrary multi-scale wavelet transform (WT) and obtained each high-dimensional wavelet coefficient matrix. The equivalence relation in time domain between wavelet coefficients and signal energy was founded. The energy eigenvalue was used as the input feature vectors of artificial neural network (ANN) for classification. This greatly decreased the number of input vectors of ANN, and the recognition rate was 89.75%. The research findings show that the feature extraction method of lung sounds is highly efficient.

The character recognition is difficulty in the Vehicle License Plate Recognition system. Characteristic dot matching, wavelet and neural networks are used in the algorithm. The characteristic dot matching is used in the number and letter recognition, but wavelet transform and neural networks are used in the Chinese. This algorithm is not only efficient in character recognition, but also may be used in fragmentary character recognition. By this algorithm, the vehicle license plate can be recognized more quickly and more correctly.

To overcome the disadvantages of dyadic wavelet and extend the scope of wavelet applications, the multiresolution analysis of the rank M wavelet and the orthogonal condition of the rank M wavelet are studied. A new method for the construction of the filter coefficients corresponding to the rank 3 wavelet scaling function is presented. And the conclusion is drawn that the number of the FIR coefficients corresponding to the rank M wavelet scaling function is at least MN, where N is the vanishing moment of the wavelet function. Rank M wavelet has potential advantages over the traditional wavelets in image compression.

Independent of the other communities who have developed theories of wavelets over the last twenty years, we developed over the same period a view of wavelets seen as stochastic processes. That context arose naturally from our theory of Time operators in statistical mechanics. Essential ingredients in our theory included Kolmogorov dynamical systems and conditional expectations. The purpose of the present paper is to come up-to-date on the relationship of our theory to the general theory of wavelets.

Snakes, also known as active contours are extensively used in computer vision and image understanding applications. They are energy minimizing deformable contours that converges at the boundary of an object in an image. Deformation in contour is caused due to internal and external forces acting on it. Internal force is derived from the contour itself and external force is invoked from the image. Traditional active contours proposed by Kass et al. only work for the normal intensity images and fail to perform segmentation task in presence of texture. This limitation comes due to the limited ability of the external force present in the traditional snake, which uses directly image pixel's intensity information for its formulation. Here, we present a new external force modeling technique for snakes, which can work in presence of texture. It uses texture features for external force modeling, which are derived using discrete wavelet transform. To demonstrate our model, we use various synthetic and natural texture images.

Face recognition is a rapidly growing research area due to it is contactless, most natural and most humanized. Many issues needed to do research exist in face recognition. A novel wavelet-based and SVM framework for face training and recognizing is proposed. FMT and PCA are also used for improving the accuracy and performance. In order for searching for the best performance, many experiments conducted with this method show that face recognition is largely affected by wavelet decomposing level, and at the same time by SVM type and kernel. So for high recognizing accuracy, proper methods and parameters are needed. The results prove that our scheme is an efficient method as 2D facial image recognition.

In this paper, a novel algorithm for numerical solutions of American option models is presented in terms of the precision integration and wavelet. We discretize the space variable employing quasi–wavelet appropriately so that the resulting semi–discrete system of equations is a system of ordinary differential equation, and then we apply precise integration method to the semi-discrete system. Numerical results show that the proposed algorithm is practicable and efficient.

Recently, Gaussian Process has been extended to classification problem and some Gaussian Process classifiers had been proposed. Engaging the Gaussian Process classifiers, in this paper, we propose a new face recognition algorithm based on wavelet face features and binary Gaussian Process classifiers. This method employs a binary tree method to extend the ability of binary Gaussian Process classifiers for dealing with multi-class problem. In the experiments, we compare our face recognition algorithm with other traditional techniques including PCA and Associative memory Model on ORL and MIT-CBCL face database. The experimental results show that our face recognition algorithm is effective and competitive.

Quantum threshold algorithm is proposed to reduce the noise of signal. Quantum superposition principle is used to construct noise model in wavelet domain. We consider that signal is quasi quantum system. Every wavelet coefficient belongs to a superposition state. We don't know whether it belongs signal or noise until we measure it. Unlike hard threshold algorithm quantum threshold algorithm hasn't a certain threshold. The probability that a wavelet coefficient belongs to signal or noise is decided by a distribution function. Finally, several experiments are made to compare the proposed method with conventional hard threshold algorithm. The pseudo-Gibbs phenomena can be reduced by this algorithm.

A fast lifted symmetrical cubic B-spline wavelet transform scheme with C² smooth is put forward in the paper for the first time, which meets the needs of face construction. Through the introduction of s-lift and w-lift operators, l filter and h filter are constructed based on the two-scale two-order vanish moment relation. A tensor-product formal two-dimensional mesh MRA subdivision and continuous level of detail control scheme is also studied to reconstruct face mesh. And a typical virtual face is constructed. The error is also analyzed and the results show that by using this LOD control model, the compression ratio is high and the distortion of facial surface is very small.

Digital audio watermarking is embedded inaudible information into digital audio data for the purposes of copyright protection, ownership verification, covert communication, and/or auxiliary data carrying. A new digital audio watermarking algorithm is proposed in this paper, which based on Neural Networks and in the Wavelet Domain. The watermark is constructed by utilizing still image and embedded into the significant wavelet coefficients of a digital audio signal. The weights of the neural networks are regulated till the relationship between the host digital audio and the watermark is approximated in the process of watermark being embedded. When compared with other algorithms, it performs watermark detection without the participation of the original digital audio and also has the advantages of transparency, low computational complexity and high robustness.

An image watermarking scheme is developed with echo-based hiding technique in the wavelet domain. In the embedding process, DWT coefficients with echo delay are utilized for watermarking. During the detection process, detect echo delay is first detected by the concept of image similarity and thus the embedded watermarks are extracted by the inverse procedure of embedding process. Experimental results confirm the watermarked image can keep high PSNR and high embedding capacity.

A human brain is the most important organ which controls the functioning of the body including heartbeat and respiration. It is an extremely complex system. Electroencephalography (EEG) is the recording of electrical activity along the scalp produced by the firing of neurons within the brain. In clinical context EEG refers to the recording of the brain's spontaneous electrical activity over a short period of time, say 20-40 minutes, as recorded from multiple electrodes placed on the scalp. The main application of EEG is in the case of epilepsy, as epileptic activity can create clear abnormalities on a standard EEG study. A secondary clinical use of EEG is in coma, Alzheimer's disease, encephalopathies, and brain death. However in the recent years EEG is also being used to design the brain of a robot. Mathematical concepts specially methods for numerical solution of partial differential equations with boundary conditions, inverse problem methods and wavelet analysis have found prominent position in the study of EEG. The present paper is devoted to this theme and will highlight the role of wavelet methods. It will also include the results obtained in our research project.

The aim of the study is to develop quantitative parameters of human electroencephalographic (EEG) recordings with epileptic seizures. We used long-lasting recordings from subjects with epilepsy obtained as part of their clinical investigation. The continuous wavelet transform of the EEG segments and the wavelet-transform modulus maxima method enable us to evaluate the energy spectra of the segments, to fin lines of local maximums, to gain the scaling exponents and to construct the singularity spectra. We have shown that the significant increase of the global energy with respect to background and the redistribution of the energy over the frequency range are observed in the patterns involving the epileptic activity. The singularity spectra expand so that the degree of inhomogenety and multifractality of the patterns enhances. Comparing the results gained for the patterns during different functional probes such as open and closed eyes or hyperventilation we demonstrate the high sensitivity of the analyzed parameters (the maximal global energy, the width and asymmetry of the singularity spectrum) for detecting the epileptic patterns.

The understanding of energy transfer mechanisms in a tokamak edge plasma is a major challenge for controlled fusion test reactors. High values of the Hurst exponent (H > 0.5) encountered in experimental probe data acquired in the scrape-off-layer (SOL) suggest the presence of long-range correlations favoring the hypothesis of an avalanche-type of radial transport. This communication aims at showing that this high value of Hurst coefficient does not necessarily imply the existence of long-time range correlations but it can be the witness of the presence of a particular behavior at small-time scales. Indeed, the development of a wavelet-based observer on synthetic signals, relying on fractional Brownian motions, has allowed the realization of a study model with mixed statistics. The associated time series, for which the H value is controlled, have been broken into blocks of variable length. Then, these different blocks have been scrambled randomly. Although potential long-range correlations have been thus destroyed, the wavelet-based estimator applied to these new synthetic signals is able to measure the original value of the Hurst parameter on a variable scale range. This approach highlights the persistence level on the scale range for several H and block size values. This technique reveals the reminiscent character of the synthetic process behavior appearing from small-time scales to long-time scales.