Showing papers by "Shu-Chuan Chu published in 2008"

Genetic Watermarking for Zerotree-Based Applications

Abstract: An optimized scheme for watermarking based on zerotrees is proposed in this paper. Digital watermarking is an effective means for protecting copyrights with three of the most fundamental requirements: watermark imperceptibility, watermark robustness, and watermark capacity. Conventional techniques in the literature mainly perform watermark embedding and extraction processes in the transform domain, including the discrete Fourier transform, discrete cosine transform, and discrete wavelet transform domains. The three watermarking requirements above are in conflict with each other; therefore, finding a way to obtain a trade-off among them is the major purpose of this paper. We first perform watermarking in the wavelet domain. Next, we properly select zerotrees in a wavelet transform with the genetic algorithm. Our simulation results not only demonstrate better performances of the watermarked images after optimization, but also reveal the robustness of the extracted watermarks under common attacks.

Self Embedding Watermarking Using Halftoning Technique

Abstract: This paper proposes a self-embedding watermarking scheme for digital images. It exploits a digital halftoning technique to transform the host image to a halftone image, in which the content features of the host image are well preserved. This halftone image is utilized as an approximate or compressed version of the host image and further adopted as a watermark. Unlike most available self-embedding watermarking techniques, in which watermark embedding, tamper detection and image recovery are block-wise-based, our scheme is a pixel-wise and block-wise composite scheme. That is, the watermark is pixel-wise permuted and embedded in the least-significant-bit plane of the host image, while in tamper detection and image recovery, a block-wise strategy is employed. This pixel-wise and block-wise complementary mechanism can overcome a common disadvantage of previous block-wise-based self-embedding watermarking techniques. Our scheme can be used for tamper detection and alteration location. Moreover, the tampered area can be approximately recovered using the extracted watermark. Experimental results demonstrate the effectiveness of the proposed scheme.

Self Embedding Watermarking Scheme Using Halftone Image

Abstract: Self embedding watermarking is a technique used for tamper detection, localization and recovery. This letter proposes a novel self embedding scheme, in which the halftone version of the host image is exploited as a watermark, instead of a JPEG-compressed version used in most existing methods. Our scheme employs a pixel-wise permuted and embedded mechanism and thus overcomes some common drawbacks of the previous methods. Experimental results demonstrate our technique is effective and practical.

Improved search strategies and extensions to k-medoids-based clustering algorithms

Abstract: In this paper two categories of improvements are suggested that can be applied to most k-medoids-based algorithms - conceptual/algorithmic improvements, and implementational improvements. These include the revisiting of the accepted cases for swap comparison and the application of partial distance searching and previous medoid indexing to clustering. Various hybrids are then applied to a number of k-medoids-based algorithms and the method is shown to be generally applicable. Experimental results on both artificial and real datasets demonstrate that when applied to CLARANS the number of distance calculations can be reduced by up to 98%.

Fusing 2DKPCA and 2D(PC)2A for Image Matrix Based Face Recognition with One Training Sample Per Person

Abstract: In the face recognition area, a so-called one sample per person problem occurred owing to the difficulties of collecting samples or storage space of systems. In this paper, we present a unified framework for image matrix based face recognition with one training sample per person. Firstly, the nonlinear and linear facial features are using proposed 2DKPCA and 2D(PC)2A method, the face images are directly used for feature extraction, and secondly a parallel fusion method is applied to fuse the facial features to construct the combined features. Experiments are implemented on three face databases to demonstrate the feasibility of proposed algorithm.

A Strategy for Opposing DC Components Watermarking

Abstract: For the watermarking technique, which embeds the watermark in the dc components of the discrete cosine transform (DCT) based image using the method presented by Huang et al. (1), we present two methods to remove the watermark from the carrier image. In our first method, it removes the watermark via varying the energies in the transform domain and redeeming the luminance in the spatial domain. And in the second method, it removes the watermark through diversifying the intensity of dc components in the transform domain. These two methods both can remove the watermark and pass the watermark detection as the images without hidden watermarks. In addition, though we remove the watermark from the watermarked image, the peak signal-to-noise ratio (PSNR) of the cleaned image, in comparison with the original carrier image, is still high.

A GA-Based Image Halftoning Technique

Abstract: In this paper, a novel halftoning technique for image is designed with genetic algorithm (GA). It is structured on the error diffusion kernel and the resulted halftone image is compared to the one which is originated from the standard error diffusion kernel with the same method by PSNR (peak signal-to-noise ratio).