Hung-Hsu Tsai

Color image watermark extraction based on support vector machines

Abstract This work proposes a novel watermarking technique called SVM-based Color Image Watermarking (SCIW), based on support vector machines (SVMs) for the authentication of color images. To protect the copyright of a color image, a signature (a watermark), which is represented by a sequence of binary data, is embedded in the color image. The watermark-extraction issue can be treated as a classification problem involving binary classes. The SCIW method constructs a set of training patterns with the use of binary labels by employing three image features, which are the differences between a local image statistic and the luminance value of the center pixel in a sliding window with three distinct shapes. This set of training patterns is gathered from a pair of images, an original image and its corresponding watermarked image in the spatial domain. A quasi-optimal hyperplane (a binary classifier) can be realized by an SVM. The SCIW method utilizes this set of training patterns to train the SVM and then applies the trained SVM to classify a set of testing patterns. Following the results produced by the classifier (the trained SVM), the SCIW method retrieves the hidden signature without the original image during watermark extraction. Experimental results have demonstrated that the SCIW method is sufficiently robust against several color-image manipulations, and that it outperforms other proposed methods considered in this work.

An SVD-based image watermarking in wavelet domain using SVR and PSO

The paper presents a novel blind watermarking scheme for image copyright protection, which is developed in the discrete wavelet transform (DWT) and is based on the singular value decomposition (SVD) and the support vector regression (SVR). Its embedding algorithm hides a watermark bit in the low-low (LL) subband of a target non-overlap block of the host image by modifying a coefficient of U component on SVD version of the block. A blind watermark-extraction is designed using a trained SVR to estimate original coefficients. Subsequently, the watermark bit can be computed using the watermarked coefficient and its corresponding estimate coefficient. Additionally, the particle swarm optimization (PSO) is further utilized to optimize the proposed scheme. Experimental results show the proposed scheme possesses significant improvements in both transparency and robustness, and is superior to existing methods under consideration here.

Robust lossless image watermarking based on α-trimmed mean algorithm and support vector machine

This paper presents a robust lossless watermarking technique, based on @a-trimmed mean algorithm and support vector machine (SVM), for image authentication. SVM is trained to memorize relationship between the watermark and the image-dependent watermark other than embedding watermark into the host image. While needing to authenticate the ownership of the image, the trained SVM is used to recover the watermark and then the recovered watermark is compared with the original watermark to determine the ownership. Meanwhile, the robustness can be enhanced using @a-trimmed mean operator against attacks. Experimental results demonstrate that the technique not only possesses the robustness to resist on image-manipulation attacks under consideration but also, in average, is superior to other existing methods being considered in the paper.

On the optimal design of fuzzy neural networks with robust learning for function approximation

A novel robust learning algorithm for optimizing fuzzy neural networks is proposed to address two important issues: how to reduce the outlier effects and how to optimize fuzzy neural networks, in the function approximation. This algorithm is able to reduce the outlier effects by cooperating with a conventional robust approach, and then to optimize fuzzy neural networks by determining the optimal learning rates which can minimize the next-step mean error at each iteration of our algorithm.

Wavelet-based image watermarking with visibility range estimation based on HVS and neural networks

This work proposes a wavelet-based image watermarking (WIW) technique, based on the human visible system (HVS) model and neural networks, for image copyright protection. A characteristic of the HVS, which is called the just noticeable difference (JND) profile, is employed in the watermark embedding to enhance the imperceptibility of the technique. First, we derive the allowable visibility ranges of the JND thresholds for all coefficients of a wavelet-transformed image. The WIW technique exploits the ranges to compute the adaptive strengths to be superimposed in the wavelet coefficients while embedding watermarks. An artificial neural network (ANN) is then used to memorize the relationships between the original wavelet coefficients and its watermark version. Consequently, the trained ANN is utilized for estimating the watermark without the original image. Many existing schemes require the original image to be involved in the calculation of the JND profile of the image. Finally, computer simulations demonstrate that both transparency and robustness of the WIW technique are superior to that of other proposed methods.

Adaptive fuzzy hybrid multichannel filters for removal of impulsive noise from color images

Abstract On the design of multichannel filters, especially in color image restoration, it is not easy to simultaneously achieve three objectives: noise attenuation, chromaticity retention, and edges or details preservation. In this paper, we propose a new class of multichannel filters called adaptive fuzzy hybrid multichannel (AFHM) filters to achieve these three objectives simultaneously. Our novel approach is mainly based on human concept (heuristic rules) and provides a significant framework to take the merits of filtering behavior of three filters: a vector median (VM) filter, a vector directional (VD) filter, and an identity filter. On the design of an AFHM filter, our approach is a powerful and flexible scheme to achieve these three objectives because human concept can be efficiently expressed by fuzzy implicative rules for improving the filtering performance. The AFHM filters are able to effectively inherit the merits of filtering behaviors of these three filters in color image restoration applications. This is the first paper to include human concept to design multichannel filters. Moreover, a faster convergence property of the learning algorithm is investigated to reduce the time complexity of the AFHM filters. Extensive simulation results illustrate that AFHM filters not only achieve these three objectives but also possess the robust and adaptive capabilities, and demonstrate that the performance of AFHM filters outperforms that of other proposed filtering techniques.

Genetic-based fuzzy hybrid multichannel filters for color image restoration

Abstract On the design of multichannel filters, especially in color image restoration, it is not easy to simultaneously achieve three objectives: noise attenuation, chromaticity retention, and edges or details preservation. In this paper we propose a new class of multichannel filters, called genetic-based fuzzy hybrid multichannel (GFHM) filters, to reach these three objectives simultaneously. The design of GFHM filters is mainly based on human concept (heuristic rules) and genetic algorithms. Because the human concept can be readily and efficiently expressed by fuzzy implicative rules, GFHM filters can take the useful characteristics of filtering behavior of three filters: a vector median, a vector directional, and an identity filter. Since genetic algorithms possess the global-searching capability for an optimal solution, they are able to effectively optimize GFHM filters to improve the filtering performance. In color image restoration applications, extensive simulation results illustrate that GFHM filters not only achieve these three objectives but also possess the robust and the adaptive capability; moreover, these simulation results also demonstrate that the performance of GFHM filters outperforms that of other proposed filtering techniques.

A zero-watermark scheme with geometrical invariants using SVM and PSO against geometrical attacks for image protection

Highlights? A zero-watermark scheme is proposed using RST invariant features, the SVM and the PSO algorithm against RST attacks for image authentication. ? The SVM-based zero-watermark scheme makes no changes to original images after embedding the owner signature of images. ? The SVM-based zero-watermark scheme requires no original image while retrieving watermarks. ? The particle swarm optimization algorithm is employed to search for a set of nearly optimal parameters of the SVM. ? In average, the SVM-based zero-watermark scheme outperforms other existing methods against RST attacks. This paper proposes a zero-watermark scheme with geometrical invariants using support vector machine (SVM) classifier against geometrical attacks for image authentication. Here geometrical attacks merely address rotation, scale, and translation (RST) operations on images. The proposed scheme is called the SVM-based zero-watermark (SZW) scheme hereafter. The SZW method makes no changes to original images while embedding the owner signature of images so as to achieve high transparency. Moreover, in order to promote the robustness to RST operations, it integrates the discrete Fourier transform (DFT) with the log-polar mapping (LPM) for finding out RST invariants of images. The SZW method then generates the secret key for a host image via performing a logical operation exclusive disjunction, an exclusive-or (XOR) operation, on the original watermark and a set of the characteristics of the RST invariants of the host image. Subsequently, a trained SVM (TSVM) is regarded as a mapping so that it can memorize the relationships between the set of characteristics of RST invariants and the secret key. During the watermark-extraction process of the SZW method, the TSVM is first fed with the set of characteristics of RST invariants of the watermarked image to get the estimated secret key. The SZW method then extracts the estimated watermark by performing the XOR operation on the set of characteristics of RST invariants and the estimated secret key. Consequently, the SZW method requires no original image while retrieving watermarks. In the paper, the particle swarm optimization (PSO) algorithm is also employed to search for a set of nearly optimal parameters of the SVM. Finally, the experimental results show that, in average, the SZW method outperforms other existing methods against RST attacks under consideration here.

Using SVM to design facial expression recognition for shape and texture features

This paper presents a novel facial emotion recognition (FER) technique, based on support vector machine (SVM), to recognize the facial emotion expression. Here it is called the FERS technique. First, a face detection method, which combines the Haar-like features (HFs) method with the self quotient image (SQI) filter, is used in the FERS technique to accurately locate the face region of an image. It can improve the detection rate due to the use of the SQI filter to overcome the insufficient light and shade light. Subsequently, angular radial transform (ART), discrete cosine transform (DCT) and Gabor filter (GF) are employed in the procedure of facial expression feature extraction. An SVM is trained and then utilized to recognize the facial expression for a queried face image. Finally, experimental results show that the recognition performance of the FERS technique can be better than that of other existing methods.

Audio Watermarking Based on HAS and Neural Networks in DCT Domain

We propose a new intelligent audio watermarking method based on the characteristics of the HAS and the techniques of neural networks in the DCT domain. The method makes the watermark imperceptible by using the audio masking characteristics of the HAS. Moreover, the method exploits a neural network for memorizing the relationships between the original audio signals and the watermarked audio signals. Therefore, the method is capable of extracting watermarks without original audio signals. Finally, the experimental results are also included to illustrate that the method significantly possesses robustness to be immune against common attacks for the copyright protection of digital audio.