Elijah Mwangi

A Geometric Attack Resistant Image Watermarking Scheme Based on Invariant Centroids

A digital image watermarking scheme that uses invariant centroids to detect and correct geometric attacks is proposed. The centroids are determined by computing a convergent geometric moment of a circular region of a predetermined size. These centroids serve as reference points for a region of interest in which a binary image is embedded using a wavelet based method and a private key to enhance robustness. In the watermark retrieval phase, the geometric attacks are estimated by identifying the centroids shift and making corrections. The centroids also help to identify the likely region in which the watermark could have been embedded. A search is performed by slightly perturbing the region of interest boundary until the highest correlation of the extracted watermark to the original watermark is obtained. Computer simulation results show that the method is effective in the retrieval of watermarks after geometric attacks. It is also robust against common signal processing attacks.

A spectral subtraction method for noise reduction in speech signals

A noise suppression algorithm for enhancing speech signals corrupted by additive white Gaussian noise is presented in this paper. The algorithm is based on the subtraction of estimated noise spectrum from the noisy speech spectrum. The performance of the algorithm is assessed by comparing the original clean speech and processed speech. An objective measure based on the log area ratios of a speech segment is used.

A wavelet based image watermarking scheme with a CDMA/Hadamard embedding technique

A new blind digital image watermarking scheme that employs a Hadamard matrix to spread the CDMA watermark in selected DWT coefficients in a gray-level cover image is proposed. The coefficients are obtained from the 1- level DWT Haar decomposition of the image. A binary image is used as a watermark and is embedded to produce a watermarked image that is perceptually indistinguishable from the original image. A secret key is used to detect the presence of the watermark in a suspect image. The experimental results obtained by computer simulation show that the proposed scheme is robust against common attacks such as additive Gaussian noise, cropping, low pass filtering, and JPEG compression.

An image watermarking method based on the singular value decomposition and the wavelet transform

A robust image watermarking scheme in which a binary image is embedded in the singular values of selected DWT blocks in the horizontal and vertical sub-bands of a 1-level decomposition of a gray-scale image is proposed. The embedded blocks are selected by a secret key to enhance imperceptibility. A watermarked image that is perceptually indistinguishable from the original is obtained. The watermarking retrieval is non-blind and requires the use of parameters extracted during the watermarking process. The performance of the proposed algorithm is tested by comparing the retrieved watermark to the original watermark. Computer simulation results show that the algorithm is robust to common signal processing attacks such as Gaussian noise, cropping, and low pass filtering. It is also resistant to JPEG compression.

A fast procedure for acquisition and reconstruction of magnetic resonance images using compressive sampling

This paper proposes a fast and robust procedure for sensing and reconstruction of sparse or compressible magnetic resonance images based on the compressive sampling theory. The algorithm starts with incoherent undersampling of the k-space data of the image using a random matrix. The undersampled data is sparsified using Haar transformation. The Haar transform coefficients of the k-space data are then reconstructed using the orthogonal matching Pursuit algorithm. The reconstructed coefficients are inverse transformed into k-space data and then into the image in spatial domain. Finally, a median filter is used to suppress the recovery noise artifacts. Experimental results show that the proposed procedure greatly reduces the image data acquisition time without significantly reducing the image quality. The results also show that the error in the reconstructed image is reduced by median filtering.

A pentomino-based path inspired demosaicking technique for the bayer color filter array

Spatial demosaicking methods are attractive due to their robust reconstruction capabilities and low computational load. This paper presents a new gradient-based spatial demosaicking algorithm with contributions from combinatorial geometry. It introduces the use of square tiling primitives called pentominoes to dictate the difference-term pairs combinations for gradient generation. The proposed algorithm is compared with current established methods. Reconstruction performance is objectively measured using three commonly documented metrics (MSE, SSIM and CPSNR). Three different image sets are used (Kodak, McMaster-IMAX and Condat) to evaluate demosaicking robustness. An significant improvement in performance over existing methods has been obtained through MATLAB simulation.

A robust image watermarking scheme invariant to rotation, scaling and translation attacks

A robust digital image watermarking scheme that is resistant to both signal processing and geometric attacks is proposed. The watermark is embedded in the Discrete Cosine Transform domain in a spread-spectrum format and Vector Quantization techniques used to compress the image. The recovery of the watermark after rotation, scaling and translation attacks is done by using Harris corner detector-based feature-points to get a Delaunay tessellation which is used to reverse the attacks. In situations where RST attacks lead to formation of substantial dark areas on the image, some reference feature-points are lost and recovered watermark is poor or entirely lost. In our proposed scheme, a procedure of estimating the RST attacks is employed by taking an average of selected triangles in the tessellation. Computer simulation results using MATLAB have been used to confirm the accuracy of our proposed scheme.

A robust magnetic resonance imaging method based on compressive sampling and clustering of sparsifying coefficients

This paper presents a novel and robust method for medical Magnetic Resonance Imaging (MRI). The proposed method utilizes the sparsity as well as clustering of the image coefficients in the wavelet transform sparsifying domain. The method shows better immunity to reconstruction noise than other Compressive Sampling (CS) based techniques. The algorithm starts with undersampling of the k-space data of the image using a random matrix followed by reconstruction of the Haar transform coefficients of the k-space data using the Orthogonal Matching Pursuit (OMP) algorithm. The transform coefficients are then modulated by a raised-cosine shaping vector that suppresses noisy artifacts in the coefficients to restore the clustering. The shaped coefficients are then transformed into k-space data. The k-space data is finally transformed into the image in spatial domain. Experimental results show that the proposed procedure gives better results than other conventional methods in terms of terms of Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE).

Single image super resolution with guided back-projection and LoG sharpening

Single image super resolution requires approximation of high frequency information that was not captured in the available low resolution image. The process may result in an image that differs significantly from the original scene if no constraints are imposed. Iterative back-projection is one method used to guide the resolution enhancement process. This paper augments the iterative back-projection with edge sharpening of the error image and guided filtering to improve the high frequency content of the image. The algorithm is tested on 27 RGB colour images including the 24 images from the Kodak lossless true colour image set. The results are compared to those obtained using bicubic interpolation, wavelet zero padding and the established edge guided method. In the average results from the 27 images, the proposed method is observed to have an improvement of 0.9% over bicubic interpolation, 5.8% over wavelet zero padding and 7.5% over new edge directed Interpolation in terms of peak signal to noise ratio. The corresponding improvement in terms of structural similarity index are 1.1%, 3% and 6.8% respectively. In addition, the proposed method has the effect of suppressing spurious colours in the enhanced image.

Large Scale Fading Pre-coder for Massive MIMO Without Cell Cooperation

In massive MIMO technology the channel is estimated using uplink training by sending an orthogonal pilot sequence from users to the base station. These sequences are re-used in the cell and also outside the cell. This gives rise to a channel estimation error referred to as pilot contamination. Large scale fading precoding which is based on the cooperation between cells has been proposed to mitigate pilot contamination. However this approach is known to limit in data transmission rate. In this paper, we propose a novel uplink training scheme to mitigate pilot contamination using a large scale fading precoding without the need of cooperation between cells. This achieves a higher transmission rate over existing method. The simulation results show that the proposed scheme improves 5% outage rate 10 times, over the existing method.