Waleed Al-Nuaimy

Automatic detection of buried utilities and solid objects with GPR using neural networks and pattern recognition

The task of locating buried utilities using ground penetrating radar is addressed, and a novel processing technique computationally suitable for on-site imaging is proposed. The developed system comprises a neural network classifier, a pattern recognition stage, and additional pre-processing, feature-extraction and image processing stages. Automatic selection of the areas of the radargram containing useful information results in a reduced data set and hence a reduction in computation time. A backpropagation neural network is employed to identify portions of the radar image corresponding to target reflections by training it to recognise the Welch power spectral density estimate of signal segments reflected from various types of buried target. This results in a classification of the radargram into useful and redundant sections, and further processing is performed only on the former. The Hough Transform is then applied to the edges of these reflections, in order to accurately identify the depth and position of the buried targets. This allows a high resolution reconstruction of the subsurface with reduced computation time. The system was tested on data containing pipes, cables and anti-personnel landmines, and the results indicate that automatic and effective detection and mapping of such structures can be achieved in near real-time.

Automated localisation of retinal optic disk using Hough transform

The retinal fundus photograph is widely used in the diagnosis and treatment of various eye diseases such as diabetic retinopathy and glaucoma. Medical image analysis and processing has great significance in the field of medicine, especially in non-invasive treatment and clinical study. Normally fundus images are manually graded by specially trained clinicians in a time-consuming and resource-intensive process. A computer-aided fundus image analysis could provide an immediate detection and characterisation of retinal features prior to specialist inspection. This paper describes a novel method to automatically localise one such feature: the optic disk. The proposed method consists of two steps: in the first step, a circular region of interest is found by first isolating the brightest area in the image by means of morphological processing, and in the second step, the Hough transform is used to detect the main circular feature (corresponding to the optical disk) within the positive horizontal gradient image within this region of interest. Initial results on a database of fundus images show that the proposed method is effective and favourable in relation to comparable techniques.

An SVD audio watermarking approach using chaotic encrypted images

This paper presents a new approach for audio watermarking using the Singular Value Decomposition (SVD) mathematical technique. The proposed approach can be used for data hiding in the audio signals transmitted over wireless networks and for multi-level security systems as will be stated in the applications section. This approach is based on embedding a chaotic encrypted watermark in the singular values of the audio signal after transforming it into a 2-D format. The selection of the chaotic encryption algorithm for watermark encryption is attributed to its permutation nature, which resists noise, filtering, and compression attacks. After watermark embedding, the audio signal is transformed again into a 1-D format. The transformation between the 1-D and 2-D formats is performed in the well-known lexicographic ordering method used in image processing. The proposed approach can be implemented on the audio signal as a whole or on a segment-by-segment basis. The segment-by-segment implementation allows embedding the same watermark several times in the audio signal, which enhances the detectability of the watermark in the presence of severe attacks. Experimental results show that the proposed audio watermarking approach maintains the high quality of the audio signal and that the watermark extraction and decryption are possible even in the presence of attacks.

Detection of exudates in retinal images using a pure splitting technique

Diabetic retinopathy is a major cause of blindness. Earliest signs of diabetic retinopathy are damage to blood vessels in the eye and then the formation of lesions in the retina. This paper presents an automated method for the detection of bright lesions (exudates) in retinal images. In this work, an adaptive thresholding based on a novel algorithm for pure splitting of the image is proposed. A coarse segmentation based on the calculation of a local variation for all image pixels is used to outline the boundaries of all candidates which have clear borders. A morphological operation is used to refine the adaptive thresholding results based on the coarse segmentation results. Using a clinician reference standard (ground truth), images with exudates were detected with 91.2% sensitivity, 99.3% specificity, and 99.5% accuracy. Due to its results the proposed method can achieve superior performance compared to existing techniques and is robust to image quality variability.

Speech enhancement with an adaptive Wiener filter

This paper proposes an adaptive Wiener filtering method for speech enhancement. This method depends on the adaptation of the filter transfer function from sample to sample based on the speech signal statistics; the local mean and the local variance. It is implemented in the time domain rather than in the frequency domain to accommodate for the time-varying nature of the speech signals. The proposed method is compared to the traditional frequency-domain Wiener filtering, spectral subtraction and wavelet denoising methods using different speech quality metrics. The simulation results reveal the superiority of the proposed Wiener filtering method in the case of Additive White Gaussian Noise (AWGN) as well as colored noise.

Background removal from GPR data using Eigenvalues

A GPR radargram of an underground scan has reflections not only from the target but many unwanted objects known as clutters. Additionally, the signal is corrupted by the direct wave and coupling effect of the antennas and background noise. In order to successfully extract the target signature, these extra noise effects need to be eliminated. Though the clutters cannot be totally removed from the data, background removal techniques suppress their effect to quite an extent. Usually mean subtraction is used as a background removal technique but the results are just satisfactory and further improvements can be made. In this paper an Eigenvalue based background removal technique in collaboration with mean subtraction is presented. This proposed method decreases the effect of clutters and the output is much more refined. Even though this method takes slightly more time than the traditional background removal methods, the output eliminates major portion of the clutter therefore the segmentation and classification stages in an automated GPR data processing system would be much more efficient hence reducing the overall time consumption for near real time GPR data processing. The method has been implemented on a number of different data sets and the results indicate that the proposed method gives significant improvement in background removal over the existing background removal methods.

Detection of landmines and underground utilities from acoustic and GPR images with a cepstral approach

This paper introduces a cepstral approach for the automatic detection of landmines and underground utilities from acoustic and ground penetrating radar (GPR) images. This approach is based on treating the problem as a pattern recognition problem. Cepstral features are extracted from a group of images, which are transformed first to 1-D signals by lexicographic ordering. Mel-frequency cepstral coefficients (MFCCs) and polynomial shape coefficients are extracted from these 1-D signals to form a database of features, which can be used to train a neural network with these features. The target detection can be performed by extracting features from any new image with the same method used in the training phase. These features are tested with the neural network to decide whether a target exists or not. The different domains are tested and compared for efficient feature extraction from the lexicographically ordered 1-D signals. Experimental results show the success of the proposed cepstral approach for landmine detection from both acoustic and GPR images at low as well as high signal to noise ratios (SNRs). Results also show that the discrete cosine transform (DCT) is the most appropriate domain for feature extraction.

Automated detection of red lesions from digital colour fundus photographs

Earliest signs of diabetic retinopathy, the major cause of vision loss, are damage to the blood vessels and the formation of lesions in the retina. Early detection of diabetic retinopathy is essential for the prevention of blindness. In this paper we present a computer-aided system to automatically identify red lesions from retinal fundus photographs. After pre-processing, a morphological technique was used to segment red lesion candidates from the background and other retinal structures. Then a rule-based classifier was used to discriminate actual red lesions from artifacts. A novel method for blood vessel detection is also proposed to refine the detection of red lesions. For a standarised test set of 219 images, the proposed method can detect red lesions with a sensitivity of 89.7% and a specificity of 98.6% (at lesion level). The performance of the proposed method shows considerable promise for detection of red lesions as well as other types of lesions.

An improved calibration technique for free-space measurement of complex permittivity

A calibration technique for free-space measurement of the permittivity and conductivity has been developed, taking account of antenna-medium coupling. Thus, the accuracy is improved compared to previously used calibration. Two examples are given to illustrate this improvement.

Automated detection of diabetic retinopathy in blurred digital fundus images

Diabetic retinopathy (DR) is a condition where the retina is damaged due to fluid leaking from the blood vessels into the retina. In extreme cases, the patient will become blind. Therefore, early detection of diabetic retinopathy is crucial to prevent blindness. The main stages of diabetic retinopathy are non-proliferate diabetes retinopathy (NPDR) and proliferate diabetes retinopathy (PDR). In this study, we propose a system for automated classification of normal, and abnormal retinal images by automatically detecting the blood vessels, hard exudates microaneurysms, entropy and homogeneity. The objective measurements such as blood vessels area, exudates area, microaneurysms area, entropy and homogeneity are computed from the processed retinal images. These objective measurements are finally fed to the artificial neural network (ANN) classifier for the automatic classification. Different approaches for image restoration are tested and compared on Fundus images. The effect of restoration on the automatic detection process is investigated in this paper.