A multiscale empirical modeling framework for system identification

A multiscale system identification methodology is presented and discussed, that extends, in a systematic way, the classical board of single-scale system identification tools to a multiscale context. The proposed approach is built upon a wavelet-based multiscale decomposition in a receding horizon sliding window that always includes the last measured values, in order to make it adequate for on-line use. Several examples are presented that illustrate different features of the multiscale modeling framework, such as its improved ability to perform prediction in output variables having most of its energy concentrated at intermediate or coarser time scales when compared to input variables, and its intrinsic smoothing capability.     

基于分块变换-融合滤波的改进EZW算法

针对EZW算法变换和编码不能同时进行,因而不能实现完整意义上的子图像编码,以及该算法在全图像的小波变换未完成前,算法不能被执行,从而带来算法复杂度较高等问题,通过对EZW算法的深入分析,从算法的变换阶段入手,提出一种基于分块变换．融合滤波的改进EZW算法。实验结果证明,该改进算法在保持EZW算法主要优点的同时,能较好地实现完整意义上的子图像编码,同时可以大幅度降低计算复杂度。     

Development of a multi-resolution framework for NUBS

The piecewise polynomial B-spline representation is a flexible tool in Computer Aided Geometric Design (CAGD) for representing and designing the geometric objects. In the field of Computer Graphics (CG), Computer Aided Design (CAD), or Computer Aided Engineering (CAE), a very useful property for a given spline model is to have locally supported basis functions. This allows localized modification of the shape. Unfortunately this property can also become a serious disadvantage when the user wishes to edit the global shape of a complex object. A multi-resolution representation is proposed as a solution to alleviate this problem.In this work, we propose a multi-resolution representation for Non-uniform B-splines (NUBS). The proposed multi-resolution model has three features that it uses control point decimation strategy for decomposing NUBS curves and it is efficient in both time and space utilization. A comparative study of the proposed work is also made with an alternate approach in the literature, which is based upon knot decimation.     

Improved scene classification using efficient low-level features and semantic cues

Prior research in scene classification has focused on mapping a set of classic low-level vision features to semantically meaningful categories using a classifier engine. In this paper, we propose improving the established paradigm by using a simplified low-level feature set to predict multiple semantic scene attributes that are integrated probabilistically to obtain a final indoor/outdoor scene classification. An initial indoor/outdoor prediction is obtained by classifying computationally efficient, low-dimensional color and wavelet texture features using support vector machines. Similar low-level features can also be used to explicitly predict the presence of semantic features including grass and sky. The semantic scene attributes are then integrated using a Bayesian network designed for improved indoor/outdoor scene classification.     

A unifying framework for lossless and progressive image coding

Progressive coding is a desirable feature for image database telebrowsing or image transmissions over low bandwidth channels. Furthermore, for some applications, exact image reconstruction is required. In this paper, we show that most of the lossless and progressive coders can be described by a common nonlinear subband decomposition scheme. This unifying framework provides useful guidelines for the analysis and improvement of the considered decomposition methods. Finally, we compare the respective performances of these methods in terms of Shannon entropy for several images and also evaluate their compression ability when combined with a hierarchical coding technique.     

因子4的正交对称尺度函数构造

本文给出尺度因子 a=4时,紧支撑对称正交尺度函数的构造,给出了算法与算例。     

Characterization of corrosion processes by current noise wavelet-based fractal and correlation analysis

Electrochemical noise data in the presence of pitting, general corrosion and passivity were analyzed using the discrete wavelet transform. The registered current noise was decomposed into a set of band-limited details, which contain information about corrosion events occurring at a determined time-scale. It has been observed that the signal variance and variances of details depend on the intensity of processes. Distribution of the signal energy among different details was characteristic for the particular type of corrosion. The characterization of corrosion processes on the basis of in the wavelet domain calculated Hurst parameter H and fractal dimension, D, of electrochemical noise signals has been established. It is concluded that general corrosion is a stationary random process with a weak persistence and D = 2.14, whereas pitting corrosion is a non-stationary process with a long memory effect and D = 1.07. Passivity is a non-stationary process near to the Brownian motion with D = 1.56. The persistence features of electrochemical noise signals were explained also by correlation coefficients calculated between signals obtained by discrete wavelet multiresolution decomposition.     

Multi-scale extension of PLS algorithm for advanced on-line process monitoring

Typical process measurements are usually correlated with each other and compounded with various phenomena occurring at different time and frequency domains. To take into account this multivariate and multi-scale nature of process dynamics, a multi-scale PLS (MSPLS) algorithm combining PLS and wavelet analysis is proposed. The MSPLS first decomposes the process measurements into separated multi-scale components using on-line wavelet transform, and then the resultant multi-scale data blocks are modeled in the framework of multi-block PLS algorithm which can describe the global relationships across the entire scale blocks as well as the localized features within each sub-block at detailed resolutions. To demonstrate the feasibility of the MSPLS method, its process monitoring abilities were tested not only for the simulated data sets containing several fault scenarios but also for a real industrial data set, and compared with the monitoring abilities of the standard PLS method on the quantitative basis. The results clearly showed that the MSPLS was superior to the standard PLS for all cases especially in that it could provide additional scale-level information about the fault characteristics as well as more sensitive fault detection ability.     

A wavelet-based framework for acquired radiometric quantity representation and accurate physical rendering

In this paper, we present a framework based on a generic representation, which is able to handle most of the radiometric quantities required by global illumination software. A sparse representation in the wavelet space is built using the separation between the directional and the wavelength dependencies of such radiometric quantities. Particularly, we show how to use this representation for spectral power distribution, spectral reflectance and phase function measurements modeling. Then, we explain how the representation is useful for performing spectral rendering. On the one hand, it speeds up spectral path tracing by importance sampling to generate reflected directions and by avoiding expensive computations usually done on-the-fly. On the other hand, it allows efficient spectral photon mapping, both in terms of memory and speed. We also show how complex light emission from real luminaires can be efficiently sampled to emit photons with our numerical model.     

Wavelet analysis to characterize cluster dynamics in a circulating fluidized bed

A common hydrodynamic feature in heavily loaded circulating fluidized beds is the presence of clusters. The continuous formation and destruction of clusters strongly influences particle hold-up, pressure drop, heat transfer at the wall, and mixing. In this paper fiber optic data is analyzed using discrete wavelet analysis to characterize the dynamic behavior of clusters. Five radial positions at three different axial locations under five different operating conditions spanning three different flow regimes were analyzed using discrete wavelets. Results are summarized with respect to cluster size and frequency.