全体要旨：
I am conducting research in the field of digital watermarking, security and privacy, steganography, sound synthesis, machine learning, natural language processing, bioinformatics, image processing, and so on. In my proposed period of stay, I have a plan to present my research achievements on digital watermarking. In the first seminar, a brief overview on digital watermarking that includes the review of the recent state-of-the-art methods, the applications areas of the watermarking, challenges to propose a new method, and our proposed watermarking methods will be presented. Besides, a few audio watermarking methods will be presented using fast Fourier transform (FFT), Cartesian-polar transform (CPT), singular value decomposition (SVD), and eigenvalue decomposition (EVD) for copyright protection. In the second seminar, some of our improved audio watermarking methods will be presented using discrete cosine transform (DCT), log-polar transform (LPT), parametric slant-Hadamard transform (PSHT), exponential-log operation (ELO), singular value decomposition (SVD) and Hessenberg decomposition (HD). In the third seminar, some of our image watermarking methods will be presented using fan beam transform (FBT), discrete linear canonical transform (DLCT), Cepstrum transform (CT), radon transform (RT), QR decomposition (QRD), Jordan decomposition (JD), and 4-connected t-o’clock scrambling for copyright protection.
1st Day : 2022年9月5日(月)
題目：Advances in Digital Watermarking Based on Matrix Decomposition
要旨：In recent years, due to the rapid development of internet and multimedia technology, transmission and distribution of multimedia data have become an extremely simple task. This has become a serious threat for multimedia content owners. Thus, there is a significant interest for copyright protection of multimedia data. Digital watermarking has drawn extensive attention for copyright protection of multimedia data. It is a process of embedding watermark into the multimedia data to show authenticity and ownership. This technique has several applications such as copyright protection, content authentication, fingerprinting, broadcast monitoring, and so on. In this seminar, a brief overview on the recent state-of-the-art methods with their limitations, properties and application areas of watermarking, challenges to propose a new method, my proposed methods and their performance in comparison with state-of-the-art methods will be presented. After rigorous survey on watermarking it was observed that the conventional methods have the difficulty in obtaining a favorable tradeoff among robustness, imperceptibility, security and payload. Besides, some methods show low robustness, whereas some are less imperceptible or less secure. To address this issue, some watermarking methods were proposed where our main aim was to find the suitable embedding location in an audio or image and the embedding process that can provide high robustness, imperceptibility, security, and payload. A semi-blind method was proposed using FFT, SVD and CPT. Initially, the watermark image is encrypted by chaotic map to enhance confidentiality of the image. The FFT is applied to each frame and low frequency FFT coefficients are selected. SVD is applied to the selected FFT coefficients of each frame represented in a matrix form. The highest two singular values of each frame are selected. The selected singular values are assumed as the components of polar coordinate system and are transformed into the components of Cartesian coordinate system. The encrypted watermark is embedded into each of the Cartesian components of the largest two singular values of low frequency FFT coefficients of each frame of the audio. The applications of semi-blind methods in some cases are limited. A blind new method based on EVD was suggested. The original audio is divided into frames and the samples of each frame are arranged into a square matrix. EVD is applied to each of these matrices. Watermark is embedded into the largest eigenvalue of each diagonal matrix by quantization. The imperceptibility of the watermarked audio signals is evaluated by using subjective listening test and objective test. Various signal processing attacks such as noise addition, cropping, resampling, requantization, MP3 compression were performed on watermarked audios to assess the robustness of the proposed methods. Simulation results indicated that the proposed methods are highly imperceptible, secure, and robust against various attacks. Moreover, they outperform state-of-the-art methods and achieve a good tradeoff among imperceptibility, robustness, security and data payload. These results verified that the proposed methods can be the suitable candidate for audio copyright protection.
2nd Day : 2022年9月12日(月)
題目：Audio Watermarking in Transform Domain Based on Singular Value Decomposition
要旨：In this seminar, audio watermarking methods using DCT, LPT, ELO, PSHT, SVD, and HD will be presented. A semi-blind method using DCT, LPT, SVD and entropy was introduced. Initially, we use a tent map that contains the chaotic characteristics to encrypt the binary watermark image for enhancing the confidentiality. The original audio is then segmented into M×M non-overlapping frames and DCT is applied to each frame. First L low frequency DCT coefficients are selected and divided into m number of sub-bands and entropy of each sub-band is calculated. The sub-band with highest entropy value is selected and SVD is applied to it. The largest singular value of the singular matrix is selected and LPT is applied to it. Watermark data is embedded into the LPT components of the largest singular value obtained from the DCT sub-band with highest entropy value of each frame by quantization. Besides, another method is proposed using DCT, SVD, ELO, and quantization. Initially, a Gaussian map is used that contains the chaotic characteristics to encrypt the binary watermark image for enhancing the confidentiality. The original audio is then segmented into non-overlapping frames and DCT is applied to each frame. Low frequency DCT coefficients are divided into sub-bands and power of each sub-band is calculated. EO is performed on the sub-band with highest power of the DCT coefficients of each frame. SVD is applied to the exponential coefficients of each sub-band with highest power represented in matrix form. Watermark is embedded into the largest singular value of the exponential component of the DCT sub-band with highest power of each frame by using a quantization function. In addition, an improved method is introduced based on PSHT and HD, where watermark is inserted into the largest value of the HD matrix of PSHT coefficients of each frame. In this algorithm, at first watermark image is preprocessed to enhance the security. Then, host signal is divided into non-overlapping frames and the samples of each frame are reshaped into a square matrix. Next, PSHT is performed on each square matrix individually and a part of this transformed matrix of size m×m is selected and HD is applied to it. Euclidean normalization is calculated from the 1st column of the Hessenberg matrix, which is further used for embedding and extracting the watermark. Simulation results ensure the imperceptibility of the proposed methods for watermarked audios. Moreover, it is demonstrated that the proposed algorithms are highly robust against numerous attacks. Furthermore, comparative analysis substantiates their superiority among other state-of-the-art methods. These results demonstrated that the proposed methods can be used efficiently for audio copyright protection.
3rd Day : 2022年9月26日(月)
題目：Image Watermarking in Transform Domain Based on Matrix Decomposition
要旨：In this seminar, image watermarking methods using RT, FBT, DLCT, CT, JD, QRD, and 4-connected t-o’clock scrambling will be presented. A color image watermarking method based on RT and JD is proposed. Initially, the host color image is converted into L*a*b* color space. Then, the b* channel is selected and it is divided into m×m non-overlapping blocks. RT is applied to each of these blocks. JD is applied to the selected RT coefficients of each block represented in p×p matrix. Watermark data is embedded in the coefficients of the similarity transform matrix of RT coefficients of the b* channel using a new quantization equation. Besides, an image watermarking method in canonical and cepstrum domains based on 4-connected t-o’clock scrambling is proposed. Initially, the watermark image is scrambled using the 4-connected t-o’clock method to enhance the security. Then the rotation operation is applied to the host image to extract the region where the watermark bits are embedded. After that, DLCT is applied to the extracted region to obtain the DLCT region. CT is performed on DLCT region to obtain CT region. The CT region is then divided into non-overlapping blocks. The watermark bits are inserted into each block using max-heap and min-heap tree property. Moreover, another blind algorithm was presented using FBT and QRD. Initially, the original image is transferred from RGB to L*a*b* color model and FBT is applied to b* component to obtain the FBT coefficients. Then FBT coefficients of the b*component are divided into m×m non-overlapping blocks and the determinant of each block is calculated. Then n blocks with larger determinants are selected and QRD is conducted to each selected block. Watermark data is placed into the first row fourth column element of the upper triangular matrix of the selected blocks with larger determinants using a new embedding function. To assess the imperceptibility, peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) of the watermarked images were calculated. Various attacks such as noise addition, cropping, JPEG compression, sharpening, filtering, contrast adjustment and so on were applied to the watermarked images to measure the robustness of our proposed methods. Experimental results demonstrated that proposed methods show high robustness against numerous attacks. Moreover, they produce high quality watermarked images and provide high security. Furthermore, they have superior performance than recent methods in terms of imperceptibility, robustness, and security. These results indicated that the proposed methods can be effectively utilized for image copyright protection.