Using artificial testing efficiently is technically difficult because of the large amounts of data which must be processed.So,according to the feature and correlation of aurora time-series image,this paper proposed an algorithm based on image segmentation to extract region of interest(ROI) of change.The analysis started with a feature extraction of the input sequence from the spatial domain.Then,it considered correlation between images in a close time sequence,proposed discrete wavelet transform(DWT) to analyze the correlation for the sake of their representative.It proposed K-means clustering to select training samples,and used feature-scaling kernel Fisher discriminant analysis(FS-KFDA) which was a modified kernel Fisher discriminant analysis to train and build classifiers to extract ROI base on the training samples.Experiments carried out on the real aurora image database from Chinese Arctic Yellowriver station point out the effectiveness of the proposed algorithm,which results in an increase of segmentation precision with respect to conventional algorithms.

Aurora is the typical ionosphere track generated by the interaction of solar wind and magnetosphere. This paper proposed a method based on eigenvalue-scaling kernel fisher discriminant analysis and Karhunen-Loeve Transform (KLT) to take advantage of interband correlation between aurora images to detect the change of aurora in serial time. Conventional classification algorithms are incapable of attaining the desired classification accuracy, and the feature of redundancy of images in a close time sequence is not considered. To solve the problems, we first apply KLT to reduce the spectral redundancies and wavelet transform to extract eigenvalues of the spatial domain. Then we employ eigenvalue-scaling kernel fisher discriminant analysis which is a modified kernel fisher discriminant analysis to realize the desired classification accuracy. Experiments carry out on time series image pointed out the effectiveness of the presented technique, which results in an increase of the classification accuracy with respect to conventional algorithms.

A lossless compression method for 3-D color images is proposed based on Improved Integer Karhunen-Loeve Transform (IIntKLT), incorporated with SPECK for color image compression. First, the IIntKLT is applied to reduce the redundancies between the color components, so that the complete reversible transform is guaranteed, and then by using SPECK, the performance of coding is improved. The results of experimentation of standard testing color image, show that the new approach has an increase of 0.1 bpp in lossless image compression, and in the case of complete reversible lossy compression, compared with JPEG2000, it at most has a increase of 0.88 dB.

A new image compression algorithm based on lifting directionlet transform (LDT) is proposed.This transform captures the multi-directional anisotropic image features effciently and processes the structure of lattice-based separable filtering and sampling.The quad-tree segmentation is designed for direction optimization of local region,and a setpartitioned embedded block algorithm for the statistic distribution property of transform coeficients is adopted.The coding performance is improved by designing the new chained list sorting and context-based arithmetic coder.The experimental results show that our proposed compression algorithm outperforms the standard wavelet-based SPECK,SPIHT,JPEG2000 and original directionlet-based methods both in terms of peak signal to noise ratio (PSNR) and visual quality.Especially at the low-rate,our algorithm can preserve better the detailed information.

In this paper,a novel method based on spectral clustering ensemble using nonnegative matrix factorization(NMF) is proposed for the segmentation of SAR image.Firstly,diversity segmentation components are obtained due to the spectral clustering method is sensitive to the scaling parameter.Secondly,these components are combined by using NMF,NMF is a method that can obtain a representation of data full of intuitive meaning and physical interpretation.Finally,segmentation result is obtained according to the combined result.To show the effectiveness of the novel method,experiments with three texture images and four SAR images are considered.The segmentation results are evaluated by comparing with K-means method,spectral clustering method based on Nystrom approximation and Meta-clustering method.According to the qualitative and quantitative analysis,the proposed method is effective and has some practical value.

More and more difficulties are encountered when using Support Vector Machine (SVM) and kernel fisher discriminant analysis (KFDA) to classify high dimension multi-spectral image. In order to overcome these limitations, feature scaling for kernel fisher discriminant analysis (ES-KFDA) is used to classify the high dimension image. The high complexity computation of SVM is mitigated greatly and the performance of KFD in the presence of many irrelevant features is improved. Incorporated with wavelet transform in space domain and analysis of characteristics between spectra, we focus here on tuning the scaling factors of the feature scaling kernel by feature scaling for kernel fisher discriminant analysis, in a feature-scaling kernel, each feature has its own scaling factor. If some feature is insignicant or irrelevant for classication, the associated scaling factor will be set smaller; otherwise, it will be set larger. So the performance of ES-KFDA in the presence of many irrelevant features is mitigated greatly. Experiment results of AVIRIS 92AV3C show that the generalization ability of ES-KFDA is strong, and its classification accuracy is better than traditional algorithms, the training time of it is significantly shorter than SVM. Copyright © 2011 Binary Information Press.

In this paper, a new lifting scheme of directionlet transform(LDT) is presented, the corresponding multi-directional and anisotropic transform has lattice-based separable filtering and subsampling along any two directions with rational slopes. We design an adaptive compression algorithm based on LDT, using the quad-tree segmentation resulting optimized directions. Experimental results show that our proposed compression algorithm for image coding outperforms the standard wavelet-based SPIHT and JPEG2000 both in terms of PSNR and visual quality, especially at the low-rate.