Showing papers by "Jeng-Shyang Pan published in 1998"

A modified tabu search approach for texture segmentation using 2-D non-separable wavelet frames

Abstract: The paper proposes a new feature vector which is characterized by a density of 2D overcomplete wavelet transform extrema estimated at the output of the corresponding filter bank and forms a feature vector for clustering. We formulated the texture segmentation problem as a combinatorial optimization. The good texture discrimination ability of the feature is demonstrated with the three-category texture image via a modified tabu search approach. According to the proposed schedule, the trial solution in this search uses the centroid of the cluster as a string and has been performed to make the objective function better in the hope that it eventually will achieve a better solution. A quantitative calculation of the accuracy of our segmentation results is presented.

Comparison study on VQ codevector index assignment.

Abstract: Vector quantization is a popular technique in low bit rate codingof speech signal. The transmission index of the codevector ishighly sensitive to channel noise. The channel distortion canbe reduced by organizing the codevector indices suitably.Several index assignment algorithms are studied comparatively.Among them, the index allocation algorithm proposed by Wuand Barba is the fastest method but the channel distortion is theworst one. The proposed parallel tabu search algorithm reachthe best performance of channel distortion. 1.INTRODUCTION Vector quantization (VQ) [1] is a widely used technique for datacompression. The binary indices of the optimally chosencodevectors are sent to the destination. A vectorXxx x={, , , } 12  k consisting of k samples of informationsource in the k-dimensional Euclidean space R k is sent to thevector quantizer. The k-dimensional vector quantizer with thenumber of codevectors N is defined as follows by using thereproduction alphabet consisting of N codevectors,Ccc c={,,, }

Wavelet-based signal approximation with multilevel learning algorithms using genetic neuron selection

Abstract: Neural networks based on wavelets are constructed to study the function learning problems. Two types of learning algorithms, the overall multilevel learning (OML) and the pyramidal multilevel learning (PML) with genetic neuron selection are comparatively studied for the convergence rate and accuracy using data samples of a piecewise defined signal. Moreover, the two algorithms are examined using orthogonal and non orthogonal bases. Experimental studies exhibit that the string representation of genetic algorithms (GA) is a key issue in determining the suitable network structures and the performances of function approximation for the two learning algorithms.

2-D Non-Separable Wavelet Bases for Texture Classification with Genetic Feature Selection.

Abstract: In this paper. the performances of testure classification based on pyramidal and uniform decomposition are comparatively studied with and without feature selection. This comparison using the subband variance as feature explores the dependence among features. It is shown that the main problem when employing 2-D non-separable wavelet transforms for testure classification is the determination of the suitable features and that yields the best classification results. A Mas-Mas algorithm which is a novel evaluation function based on genetic algorithms is presented to evaluate the classification performance of each subset of selected features. Esperimental results have shown the selectivity of the proposed approach and do capture the testure characteristics.