Qin Qianqing

Hierarchical support vector machines

The speed and accuracy of a hierarchical SVM (H-SVM) depend on its tree structure. To achieve high performance, more separable classes should be separated at the upper nodes of a decision tree. Because SVM separates classes at feature space determined by the kernel function, separability in feature space should be considered. In this paper, a separability measure in feature space based on support vector data description is proposed. Based on this measure, we present two kinds of H-SVM, binary tree SVM and k-tree SVM, the decision trees of which are constructed with two bottom-up agglomerative clustering algorithms respectively. Results of experimentation with remotely sensed data validate the effectiveness of the two proposed H-SVM.

Multi-focus Images Fusion Based on Data Assimilation and Genetic Algorithm

The traditional fusion algorithms, such as principal component analysis, wavelet transform, Gauss-Laplacian pyramids, Brovey transform, curvelet transform and so on, set down the fusion rules before fusion process. However, the rules which determine the attributes of fusion results cannot be adjusted according to different application. In this paper, a framework based on data assimilation and genetic algorithm for multi-focus image fusion is proposed. Data assimilation is to combine the observational data and simulative data to obtain more objective result which is firstly used in weather field. Under this framework, weights of different attributes according to the application are determined and object function constituted by the weighted sum of each evaluation index is constructed to obtain the proper fusion image. The experiments validate the feasibility of the framework.

Convolution Morphological Filters

In this paper, we present a novel morphological filter, called convolution morphological filters (CMF), using the linear convolution theory and method. The newly introduced filters employ a weighted convolution kernel and apply multiplication and division in place of addition and subtraction in ordinary morphological operations. The property of CMF indicates that it can smooth image and using to remove noise contained in images. An automatic generation algorithm of convolution morphological kernel is proposed, which is important to the function of CMF. Experimental results prove that an algorithm to smooth image or remove image noise is feasible and available. For some cases, the CMF act better than ordinary morphological filters.

Spread spectrum watermark based on wavelet transform for still digital image

The proposed method inserts a watermark into the spectral components of the data using techniques analogous to spread spectrum communications, hiding a narrow band signal in a wideband channel. The watermark is difficult for an attacker to remove. It is also robust to common signal and geometric distortion such as JPEG compression, cropping and scaling. In addition, the watermark can be extracted accurately without source host image at the receiver.

Single-Frame Remote Sensing Image Superresolution Reconstruction with Sub-blocks Noniterative Scheme

The satellite remote image is degraded obviously due to the diffraction-limitation of imaging system and the atmosphere turbulence and etc.. Namely, the observed image blurred by a low-pass filter whose transfer function vanishes beyond imaging system cut-off spatial frequency (omegac ).Processes that achieve the recreation of frequencies beyond the image pass- band are usually referred to as super-resolution algorithms. Although, multi-frame super-resolution has more potential for spatial resolution improvement, as for remote sensing image, single-frame super-resolution is prerequisite in many cases. One remote imaging blurred model is put forward basing on PSF (point spread function) optimal estimation. Then a sub-blocks non-iterative method is adapted to improve the image spatial resolution effectively with only single-frame low spatial resolution image. With this scheme, the high spatial resolution spots5 panchromatic 2.5 m image can be reconstructed from single-frame 5 m panchromatic spots5 image. The MTF is used as impersonal standard to estimate the spatial resolution improvement and image quality. As result, the high frequency detail information and effective band of image are compensated perfectly with SNR improved.

Lifting Scheme and Its Application to Edge Detection

Lifting scheme is a useful and very general technique for constructing wavelet decomposition. The paper adapts the lifting into redundant lifting to obtain shift invariant wavelet transform. In prediction and update stages of the lifting morphological operator is adopted for preserving local maxima of a signal over several scales, which is particularly useful in wavelet-based signal detection. The new transform presented in the paper is applied in multiresolution edge detection of medical image and experiment results are given to show better performance and applicable potentiality.

A New Multiscale Segmentation Algorithm on Texture Image in Wavelet Domain

With the illumination of the basic idea of model-based texture analysis methods, a new feature extraction method, Finite Texture Mixture Pattern (FTMP), was proposed in this paper. FTMP is a two-tuplet set, which can be obtained by the clustering methods. Firstly, the multi-scale and multi-direction variations are calculated. Secondly, these variations of each scale are clustered into groups respectively. The centers and their corresponding proportions composite FTMP, which describes the primary variations of different scales and different directions. Such a feature extraction method takes full advantage of the idea of model-based method, but avoids the complicate parameter estimation and expression computation. Based on FTMP, a supervised multi-scale texture image segmentation algorithm-FTMPseg is proposed, and its effectiveness is proven by quantitative and qualitative experiments.

Remote Sensing Image Fusion Based on Data Assimilation and Genetic Simulated Annealing Algorithm

The available remote sensing image fusion methods, such as that based on color space transform, on statistical (e.g. principle component analysis), on multi-scale analysis (e.g. pyramid decomposition, wavelet transform, etc.), basically set down the fusion rules before fusion process. The rules which determine the attributes of fusion results cannot be adjusted according to different application. In this paper, a framework based on data assimilation for multispectral and panchromatic image fusion is proposed. Data assimilation is to combine the observational data and simulative data to obtain more objective result which is firstly used in weather field. Under this framework, weights of different attributes are determined according to their importance degree to the following process and object function constituted by the weighted sum of each evaluation index is constructed. Finally, the object fusion is optimized through genetic simulated annealing to obtain the proper image. The experiments validate the feasibility of the framework.

A fast image retrieval algorithm with multi-channel textural features in PACS

The paper presents a fast algorithm for image retrieval using multi-channel textural features in medical picture archiving and communication system (PACS). By choosing different linear or nonlinear operators in prediction and update lifting step, the linear or nonlinearM-band wavelet decomposition can be achieved inM-band lifting. It provides the advantages such as fast transform, in-place calculation and integer-integer transform. The set of wavelet moment forms multi-channel textural feature vector related to the texture distribution of each wavelet images. The experimental results of CT image database show that the retrieval approach of multi-channel textural features is effective for image indexing and has lower computational complexity and less memory. It is much casier to implement in hardware and suitable for the applications of real time medical processing system.