Esteban Rodriguez-Marek

Perfect reconstructable decimated two-dimensional empirical mode decomposition filter banks

Traditional two-dimensional empirical mode decomposition (2D-EMD) algorithms generate multiple subband signals, each having the same size of the original signal. Thus, huge amounts of data to be stored may be generated. Moreover, the computational load is massive as the decomposition levels increase. This paper introduces a method to reduce the data generated (i.e. reduce storage requirement) by incorporating decimation into the 2D-EMD, while maintaining perfect reconstruction. Furthermore, it is well established that traditional EMDs can be thought as having the structure of a single dyadic filter bank. The proposed algorithm is applicable into any arbitrary tree structures including octave filter banks, 2D-EMD packets when applied to a full binary tree, etc. The methodology hereby presented builds on the algorithm introduced by the authors in [8].

A new two dimensional empirical mode decomposition for images using inpainting

This paper introduces a two dimensional version of the empirical mode decomposition (2D-EMD) using an inpainting technique. The most natural use for 2D-EMD is in image processing. The algorithm hereby presented is based on the underlining idea of the one-dimensional EMD presented by Huang et al. in [1], but using inpainting and three surfaces: upper surface, lower surface, and mean surface. Upper and lower surfaces are surfaces connected by local maxima and minima, respectively. The mean surface is calculated by the average of the upper and lower surfaces. Using these three surfaces, the intrinsic mode functions (IMFs) are obtained through an iterative process, as in the 1D-EMD. The lower IMFs of the 1D-EMD correspond to high frequency components. The lower IMFs obtained with 2D-EMD show image edges.

A Novel Data Dependent Multimedia Encryption Algorithm Secure Against Chosen-Plaintext Attacks

A novel encryption algorithm secure to chosen-plaintext attacks is presented. As opposed to traditional key algorithms, one of the keys in the algorithm presented depends on the message itself. Two encryption matrices are generated by means of singular value decomposition (SVD), using a portion of the message. The two encryption matrices generated are further multiplied into the left and right sides of other data frames for encryption in the transmitter. Without additional information, except for a key and an integer for signs, the encryption matrices can be found and, thus, the original data obtained at the receiver. This is done by exploiting special properties of the SVD of real symmetric matrices. Hence, the algorithm performs time-varying encryption (and, thus, decryption), i.e. the algorithm generates time-varying ciphertexts depending on both the design parameters and the plaintext itself. Since the encryption depends on message data, it leads to a good solution to various known attacks, including chosen-plaintext attacks. The algorithm can be applied to any signal such as text, audio, and image, etc.

A new infrared image fusion method using empirical mode decomposition and inpainting

This paper puts forward a new method to fuse infrared images using empirical mode decomposition (EMD) and inpainting algorithms. EMD is a non-parametric, data-driven analysis tool that decomposes non-linear, non-stationary signals into a set of signals denominated intrinsic mode functions (IMFs) and a residual. Fusion rules are set up to fuse the corresponding IMFs and residual by designing for the weighting factor to emphasize desirable features of the original images. The image is then reconstructed using fused IMFs and residuals. This new image fusion algorithm is evaluated based on several tests such as edge information, mutual information, and information entropy. Test results show that the proposed method is effective when fusing infrared images, as the fused images are very clear and include rich information from the original sources.

System Level Performance Assessment of SOC Processors with SystemC

This paper presents a system level methodology for modeling, and analyzing the performance of system-on-chip (SOC) processors. The solution adopted focuses on minimizing assessment time by modeling processors behavior only in terms of the performance metrics of interest. Formally, the desired behavior is captured through a C/C++ executable model, which uses finite state machines (FSM) as the underlying model of computation (MOC). To illustrate and validate our methodology we applied it to the design of a 16-bit reduced instruction set (RISC) processor. The performance metrics used to assess the quality of the design considered are power consumption and execution time. However, the methodology can be extended to any performance metric. The results obtained demonstrate the robustness of the proposed method both in terms of assessment time and accuracy

A novel data encryption algorithm based on wavelet filter banks and the singular value decomposition

We present an algorithm which performs data encryption by serially concatenating two transform stages. The outer stage uses one of the orthogonal matrices obtained from the singular value decomposition (SVD) of an arbitrary signal, such as white noise or the sum of cosines of different frequencies. The inner stage of encryption uses a fast, parallelized wavelet filter bank using our previously presented algorithm (Koh, M.S. and Rodriguez-Marek, E., Proc. IEEE Int. Symp. on Sig. Process. and Inform., 2003). This algorithm is generalized for an arbitrary number of nodes and decomposition levels. Past algorithms based on the wavelet packet tree structure present a drawback for band-limited signals, because attackers can guess the approximate frequency bands of the wavelet decomposition. Our algorithm uses orthogonal matrices generated by the SVD, which spread the frequency content of the signal into the available spectrum when applied to the original vector. Furthermore, the algorithm is based on parallelized filter banks, which provide a flexible and highly adaptive structure for encryption and decryption.

An image fusion method based on quotient singular value decomposition

This paper presents a new image fusion algorithm that combines quotient singular value decomposition (QSVD) with a simple averaging operation. Multi-focused images are first averaged into a new image. Then, error images are obtained with the averaged image and the multi-focused images. The most error-contributing component in each error image is replaced by the most contributing image component in the multi-focused image using QSVD in order to reduce errors. With each reduced error image, a new singular vector is calculated to get fused images. The final infused image is then decided by calculating the standard deviation of each fused image. Experiment results such as mutual information (MI), information entropy (IE), edge preservation information (Qabf), signal-to-noise-ratio (SNR) and root mean square error (RMSE) are used to evaluate the algorithm. The experimental results show that the developed algorithm is an efficient fusion algorithm.

Video coding over packet-erasure channels

Spatial domain redundancy is studied for packet erasure recovery in a video coding system. The image is partitioned into a checkerboard pattern of white (W) and black (B) blocks. A subset of pixels in the B (W) blocks is selected to be encoded as redundant information with the W(B) blocks. The Yu, Liu, and Marcellin (1996) approach is generalized and several different redundancy patterns are investigated. Significant improvement in interpolation performance is achieved with different redundancy patterns.