Showing papers by "Jeng-Shyang Pan published in 2004"

Constrained ant colony optimization for data clustering

Abstract: Processes that simulate natural phenomena have successfully been applied to a number of problems for which no simple mathematical solution is known or is practicable. Such meta-heuristic algorithms include genetic algorithms, particle swarm optimization and ant colony systems and have received increasing attention in recent years. This paper extends ant colony systems and discusses a novel data clustering process using Constrained Ant Colony Optimization (CACO). The CACO algorithm extends the Ant Colony Optimization algorithm by accommodating a quadratic distance metric, the Sum of K Nearest Neighbor Distances (SKNND) metric, constrained addition of pheromone and a shrinking range strategy to improve data clustering. We show that the CACO algorithm can resolve the problems of clusters with arbitrary shapes, clusters with outliers and bridges between clusters.

Robust image watermarking based on multiple description vector quantisation

Abstract: An innovative scheme for watermarking based on vector quantisation for transmitting over noisy channels is proposed. By modifying multiple description vector quantisation for watermark embedding and extraction, simulation results not only demonstrate effective transmission of the watermarked image, but reveal the robustness of the extracted watermark.

Hadamard transform based fast codeword search algorithm for high-dimensional VQ encoding

Abstract: An efficient nearest neighbor codeword search algorithm based on Hadamard transform for vector quantization is presented in This work. Four efficient elimination criteria are derived from two important inequalities based on three characteristic values in the Hadamard transform domain. Before the encoding process, all codewords in the codebook are Hadamard-transformed and sorted in the ascending order of their first elements. During the encoding process, we firstly perform the transform on the input vector and calculate its characteristic values, and initialize the current closest codeword of the input vector to be the codeword whose first element of Hadamard transform is nearest to that of the input vector, and secondly use the proposed elimination criteria to find the nearest codeword to the input vector using the up-down search mechanism near the initial best-match codeword. Experimental results demonstrate the proposed algorithm is much more efficient than most existing nearest neighbor codeword search algorithms in the case of high dimension.

Robust VQ-based digital watermarking for memoryless binary symmetric channel

Abstract: Vector quantization (VQ) has been distinguished for its high compression rate in lossy data compression applications. And VQ-based watermarking plays a newly developed branch in digital watermarking research fields. In this paper, we propose optimized schemes for VQ-based image watermarking. We overcome the VQ index assignment problem with genetic algorithm, which is suitable for transmitting the watermarked image over noisy channels. We obtain better robustness of the watermarking algorithm against the effects caused by channel noise in our simulations after inspecting the results from several test images. In addition, to compare with existing schemes in literature, the watermarked image quality in our scheme has approximately the same quality, with better performance in robustness, to the schemes proposed by other researchers. This also proves the effectiveness of our proposed schemes in VQ-based image watermarking for copyright protection.

Robust VQ-Based digital watermarking for the memoryless binary symmetric channel

Abstract: Vector quantization (VQ) has been distinguished for its high compression rate in lossy data compression applications. And VQ-based watermarking plays a newly developed branch in digital watermarking research fields. In this paper, we propose optimized schemes for VQ-based image watermarking. We overcome the VQ index assignment problem with genetic algorithm, which is suitable for transmitting the watermarked image over noisy channels. We obtain better robustness of the watermarking algorithm against the effects caused by channel noise in our simulations after inspecting the results from several test images. In addition, to compare with existing schemes in literature, the watermarked image quality in our scheme has approximately the same quality, with better performance in robustness, to the schemes proposed by other researchers. This also proves the effectiveness of our proposed schemes in VQ-based image watermarking for copyright protection.

Improved K nearest neighbor classification algorithm

Abstract: A novel and efficient algorithm is proposed to reduce the computational complexity for KNN classification. It uses two important features, the approximation coefficient of a fully decomposed feature vector with Haar wavelet and the variance of the corresponding untransformed vector, to produce two efficient test conditions. Since those vectors that are impossible to be the k closest vectors in the design set are kicked out quickly by these conditions, this algorithm saves largely the classification time and have the same classification performance as that of the exhaustive search classification algorithm. Experimental results based on texture image classification verify our proposed algorithm.

A VQ-based image-in-image data hiding scheme

Abstract: Vector quantization (VQ)-based data hiding and watermarking techniques form a newly developed branch in the digital watermarking research field. We propose a novel watermarking scheme for image-in-image data hiding based on VQ. Unlike existing schemes in the literature for embedding a binary watermark into the original image with VQ, we present an efficient and secure algorithm for embedding a gray-level watermark into the original. The offered scheme is robust to attacks such as VQ compression and JPEG compression. Moreover, it greatly expands the capacity of the watermarking system. Experimental results demonstrate the superiority of our scheme in watermark robustness and watermark capacity. The results also prove the effectiveness of our proposed schemes in VQ-based image-in-image data hiding.

VQ-Based gray watermark hiding scheme and genetic index assignment

Abstract: A vector quantisation (VQ) based watermarking scheme for hiding the gray watermark is presented. It expands the watermark size, and employs VQ index assignment procedure with genetic algorithm, called genetic index assignment (GIA), for watermarking. The gray watermark is coded by VQ, and obtained indices are translated into a binary bitstream with a much smaller size. We partition the codebook into two sub-codebooks, and use either one of them based on the value of the bit for embedding. Next, GIA is employed to find better imperceptibility of watermarked image. Experimental results show that the proposed method possesses advantages over other related methods in literature.

Improved partial distance search for k nearest-neighbor classification

Abstract: An important method in pattern recognition is k nearest-neighbor classification. However, its computational complexity limits its real-time applications. The partial distance search is a solution to this problem. Although it is not very effective, it can be combined with other algorithms to reduce the complexity. The paper proposes an indexing method that uses the variance vector of feature vectors of a design set to improve the efficiency of the partial distance search. Experimental results indicate the effectiveness of this indexing preprocessing

Vq-based gray watermark embedding scheme with genetic index assignment

Abstract: A digital watermarking scheme based on vector quantisation (VQ) for gray watermark is proposed. It begins with the procedure of encoding the gray watermark, and embeds the encoded indices into the cover image in VQ domain then. To improve the performance of the watermarking scheme, a genetic index assignment (GIA) procedure, which modifies the signal of the watermark to suit the signal of the cover image, is proposed. The proposed gray watermark embedding scheme is easy to implement, requires no original cover image to be presented during extraction, expands the size of the used watermark, and provides better watermarked results. Experimental results will show the novelty and effectiveness of it.

Design of hierarchical keys for a multi-user-based watermarking system

Abstract: For a multi-user-based watermarking system, a new scheme for providing the function of secret sharing is proposed. A user-key generating procedure is introduced to generate one master key and several normal keys. By using either of these normal keys, a secret watermark is obtained from the cover image. By referring to the original watermark and all the generated secret watermarks, a public watermark is created and embedded into the cover image. The proposed scheme does not require the original image to be presented during extracting. To reveal the genuine watermark from the watermarked image, except for the super-user who can extract it directly by using the master key, the normal users who share the secret can only achieve it by presenting the shadow watermarks extracted by using their own keys

Novel multi-centroid, multi-run sampling schemes for $K$-medoids-based algorithms

Abstract: Clustering in data mining is used to group similar objects based on their distance, connectivity, relative density, or some specific characteristics. Data clustering has become an important task for discovering significant patterns and characteristics in large spatial databases. The k-medoids-based algorithms have been shown to be effective to spherical-shaped clusters with outliers. However, they are not efficient for large database. In this paper, we propose two novel algorithms - Multi-Centroid with Multi-Run Sampling Scheme, which we termed MCMRS, and a more advanced sampling scheme termed the Incremental Multi-Centroid, Multi-Run Sampling Scheme, which called simply (IMCMRS) hereafter, to improve the performance of many k-medoids-based algorithms including PAM, CLARA and CLARANS. Experimental results demonstrate the proposed scheme can not only reduce by more than 80% computation time but also reduce the average distance per object compared with CLARA and CLARANS. IMCMRS is also superior to MCMRS.

Digital image watermarking approach based on lapped orthogonal transform

Abstract: A new watermarking approach based on lapped orthogonal transform (LOT) is proposed. It provides better imperceptibility by reducing the blocking effects which exist in most block-based image coding transforms like JPEG. The proposed scheme performs the LOT to the cover image, modifies the LOT coefficients to hide the watermark bits according to the given user-key, and executes the inverse LOT to generate a watermarked image. It requires no original cover image to be introduced during extraction, and it has strong robustness under the JPEG attacks.

New Search Strategies And A New DerivedInequality For Efficient K-medoids-basedAlgorithms

Abstract: In this paper, a new inequality is derived which can be used for the problem of nearest neighbor searching. We also present a searching technique referred to as a previous medoid index to reduce the computation time particularly for the kmedoids-based algorithms. A novel method is also proposed to reduce the computational complexity by the utilization of memory. Four new search strategies for k-medoids-based algorithms based on the new inequality, previous medoid index, the utilization of memory, triangular inequality criteria and partial distance search are proposed.Experimentalresults demonstratethatthe proposedalgorithm applied to the CLARANS algorithm may reduce the computation time from 88.8% to 95.3% with the same average distance per object comparing with CLALRANS. The derived new inequality and proposed search strategies can also be applied to the nearest neighbor searching and the other clustering algorithms.

Hadamard transform based equal-average equal-variance equal-norm nearest neighbor codeword search algorithm

Abstract: The paper presents a novel, efficient, nearest-neighbor codeword search algorithm based on three elimination criteria in the Hadamard transform (HT) domain. Before the search process, all codewords in the codebook are Hadamard-transformed and sorted in the ascending order of their first elements. During the search process, we first perform the HT on the input vector and calculate its variance and norm, and secondly exploit three efficient elimination criteria to find the nearest codeword to the input vector using the up-down search mechanism near the initial best-match codeword. Experimental results demonstrate that the performance of the proposed algorithm is much better than that of most existing nearest neighbor codeword search algorithms, especially in the case of high dimension