Cabir Vural

Blind image deconvolution via dispersion minimization

In linear image restoration, the point spread function of the degrading system is assumed known even though this information is usually not available in real applications. As a result, both blur identification and image restoration must be performed from the observed noisy blurred image. This paper presents a computationally simple iterative blind image deconvolution method which is based on non-linear adaptive filtering. The new method is applicable to minimum as well as mixed phase blurs. The noisy blurred image is assumed to be the output of a two-dimensional linear shift-invariant system with an unknown point spread function contaminated by an additive noise. The method passes the noisy blurred image through a two-dimensional finite impulse response adaptive filter whose parameters are updated by minimizing the dispersion. When convergence occurs, the adaptive filter provides an approximate inverse of the point spread function. Moreover, its output is an estimate of the unobserved true image. Experimental results are provided.

Determination of Sleep Stage Separation Ability of Features Extracted from EEG Signals Using Principle Component Analysis

In this study, a method was proposed in order to determine how well features extracted from the EEG signals for the purpose of sleep stage classification separate the sleep stages. The proposed method is based on the principle component analysis known also as the Karhunen–Loéve transform. Features frequently used in the sleep stage classification studies were divided into three main groups: (i) time-domain features, (ii) frequency-domain features, and (iii) hybrid features. That how well features in each group separate the sleep stages was determined by performing extensive simulations and it was seen that the results obtained are in agreement with those available in the literature. Considering the fact that sleep stage classification algorithms consist of two steps, namely feature extraction and classification, it will be possible to tell a priori whether the classification step will provide successful results or not without carrying out its realization thanks to the proposed method.

Blind deconvolution of noisy blurred images via dispersion minimization

In linear image restoration, the point spread function of the degrading system is assumed known even though this information is usually not available in real applications. As a result, both blur identification and image restoration must be performed from the observed noisy blurred image. This paper presents a computationally simple linear adaptive finite impulse response filter for blind image deconvolution. This is essentially a two-dimensional version of the constant modulus algorithm that is well known in the field of blind equalization. The two-dimensional extension is shown capable of reconstructing noisy blurred images using partial a priori information about the true image and the point spread function. The method is applicable to minimum as well as mixed phase blurs. Experimental results are provided.

Blind equalization of single-input single-output fir channels for chaotic communication systems

Recently we have developed a simplified recursive adaptive blind channel equalization method for Single-Input Single-Output (SISO) chaotic communication systems. Even though the simplified recursive algorithm gives superior results compared to the state of the art chaotic blind channel equalization algorithms, it has a very important limitation: convergence of the adaptive algorithm is ensured for only Strictly Positive Real (SPR) channels. In this study, we propose a non-recursive chaotic blind channel equalization algorithm that works regardless of whether the channel is SPR or not. First, a statistically optimum fixed filter is designed assuming that the channel is known. Then, it is shown via computer simulations that its performance is very close to that of the statistically optimum fixed filter. Furthermore, it gives better results especially for non-SPR channels compared to the well-known minimum nonlinear prediction error method and the simplified recursive algorithm developed in our previous work. The method is computationally simple and does not impose any restrictions on the channel other than being a finite impulse response filter. Since the instantaneous gradient is used to derive the adaptive algorithm, the proposed method works for slowly and smoothly varying linear channels as well.

Blind image resolution enhancement based on a 2D constant modulus algorithm

In almost all super-resolution methods, the blur operator is assumed to be known. However, in practical situations this operator is not available or available only to a finite extent. In this paper, a super-resolution algorithm is presented in which the assumption of availability of the blur parameters is not necessary. It is a two-dimensional and single-input multiple-output extension of the well-known constant modulus algorithm which is widely used for blind equalization in communication systems. The algorithm consists of determining a set of deconvolution filters to be applied on re-sized low-resolution and low-quality images and is suitable for pure translational motion only. An important property of the method is that the blur operators do not have to be the same for the observed low-resolution images, and also they do not need to be shift-invariant. Experimental results show that the proposed method can satisfactorily reconstruct the high-resolution image and remove the blur especially for five or less-bit images.

Recursive blind image deconvolution via dispersion minimization

This paper presents a method that uses an autoregressive filter for deblurring noisy blurred images blindly. The approach has several important advantages over using a finite impulse response filter. The optimum support of the adaptive autoregressive filter is the same as the support of the blur, and so the truncation error introduced by the finite support of the adaptive finite impulse response filter can be made arbitrarily small. Furthermore, the method can also be used for blur identification. In addition, the resulting improvement in signal-to-noise ratios are higher and convergence of the adaptive filter coefficients is faster for a given blur. First, an autoregressive method is naively derived via a gradient method to minimize the dispersion. This leads to a recursion within a recursion which is computationally complex. Next, a simplification of the method is proposed. Finally, simulations demonstrate the performance of the simplified method.

Reversible video watermarking based on histogram modification of motion compansated prediction error

In occluded areas conventional motion estimation techniques lead to high prediction errors. To reduce the prediction error, occlusion detection algorithms should be employed in motion estimation. This paper investigates the influence of the occlusion in a reversible video watermarking technique based on histogram modification of motion compensated prediction error and presents a new reversible video watermarking algorithm that has a better performance in occluded areas.

A new reversible video watermarking method based on motion compensated interpolation

Digital watermarking can be defined as embedding information into digital signals. Original signal is distorted as a result of watermarking. The goal in reversible watermaking is to reconstruct the original signal from the watermarked signal without error. In this study a new reversible watermaking algorithm based on motion compensated interpolation error is developed for digital video. The method not only provides high capacity but also creates negligible distortion in the original signal. The method is shown to be superior to the existing algorithms in terms of capacity and image quality by means of computer simulations.

Complex mapping-based blind image superresolution

Recently, we proposed a blind resolution enhancement method for pure translational motion and shift invariant blur which used a two-dimensional and single-input multiple-output extension of the constant modulus algorithm. The method worked well when the bit number per pixel was low, but the performance decreased as the bit number increased. In this work, we propose a refined scheme in which complex representation of images and a set of complex deconvolution FIR filters are used. Simulations show that the refined method succeeds in recontructing the high-resolution image without the knowledge of blur parameters even when the number of bits per pixel is high.

Robust digital image watermarking based on normalization and complex wavelet transform

In this study, a new digital image watermarking algorithm based on moment-based image normalization and the two dimensional dual tree complex wavelet transform (2D DT-CWT) was developed. Normalization provides robustness against geometrical distortions, whereas 2D DT-CWT increases robustness for attacks such as additive noise, linear and nonlinear filtering, JPEG compression. It was accomplished that added watermark achieves both transparency and robustness requirements by taking the properties of the human visual system account.