Norashikin Yahya

Subspace-based technique for speckle noise reduction in ultrasound images

Background and purposeUltrasound imaging is a very essential technique in medical diagnosis due to its being safe, economical and non-invasive nature. Despite its popularity, the US images, however, are corrupted with speckle noise, which reduces US images qualities, hampering image interpretation and processing stage. Hence, there are many efforts made by researches to formulate various despeckling methods for speckle reduction in US images.MethodsIn this paper, a subspace-based speckle reduction technique in ultrasound images is proposed. The fundamental principle of subspace-based despeckling technique is to convert multiplicative speckle noise into additive via logarithmic transformation, then to decompose the vector space of the noisy image into signal and noise subspaces. Image enhancement is achieved by nulling the noise subspace and estimating the clean image from the remaining signal subspace. Linear estimation of the clean image is derived by minimizing image distortion while maintaining the residual noise energy below some given threshold. The real US data for validation purposes were acquired under the IRB protocol (200210851-7) at the University of California Davis, which is also consistent with NIH requirements.ResultsExperiments are carried out using a synthetically generated B-mode ultrasound image, a computer generated cyst image and real ultrasound images. The performance of the proposed technique is compared with Lee, homomorphic wavelet and squeeze box filter (SBF) in terms of noise variance reduction, mean preservation, texture preservation and ultrasound despeckling assessment index (USDSAI). The results indicate better noise reduction capability with the simulated images by the SDC than Lee, Wavelet and SBF in addition to less blurry effect. With the real case scenario, the SDC, Lee, Wavelet and SBF are tested with images obtained from raw radio frequency (RF) data. Results generated using real US data indicate that, in addition to good contrast enhancement, the autocorrelation results shows better preservation of image texture by SDC than Lee, Wavelet and SBF.ConclusionIn general, the performance of the SDC filter is better than Lee, Wavelet and SBF in terms of noise reduction, improvement in image contrast and preservation of the autocorrelation profiles. Furthermore, the filter required less computational time compared to Lee, Wavelet and SBF, which indicates its suitability for real time application.

Subspace-based signature verification technique using reduced-sensor data glove

Online signature verification is a dynamic method in which the biometric system recognizes the signature by analyzing its characters such as acceleration, pressure, and orientation. The system make used of a data glove which is originally designed for virtual reality application as an input device. The proposed technique for online signature verification is based on the Singular Value Decomposition (SVD) technique which involves four aspects: 1) data acquisition and preprocessing 2) feature extraction 3) matching (classification), 4) decision making. The SVD is used to find r-singular vectors sensing the maximal energy of the signature data matrix A, called principle subspace thus account for most of the variation in the original data. Having modeled the signature through its r-th principal subspace, the authenticity of the tried signature can be determined by calculating the average distance between its principal subspace and the template signature. In this paper, we investigate the performance of the signature verification system with reduced-sensor data glove. The selection of the most prominent sensors of the data glove is based on the F-value for each sensor. The signature verification technique was tested with large number of authentic and forgery signatures and has demonstrated that it has the potential to offer high level of security for special application, such as banking and electronic commerce.

Principle subspace-based signature verification technique

Online signature verification rests on hypothesis of which for any writer, there will be similarity among signature samples, with some scale variability and small distortion. The signature verification using SVD technique is a dynamic method in which when users sign, the system recognizes the signature by analyzing its characters such as acceleration, pressure, and orientation. The SVD technique is used to find r-singular vectors sensing the maximal energy of the signature data matrix A. The data matrix is obtained using a virtual reality glove, 5DT Data Glove Ultra 14. The r-singular vector is called principle subspace which account for most of the variation in the original data. Having modeled the signature through its r-th principal subspace, the authenticity of the tried signature can be determined by calculating the average distance between its principal subspace and the template signature. The signature verification technique was tested with large number of authentic and forgery signatures and has demonstrated the good potential of this technique.

Epileptic seizure detection using singular values and classical features of EEG signals

In this paper, an epileptic seizure event detection algorithm utilizing five features namely singular values, total average power, delta band average power, variance and mean, is proposed. Using CHB-MIT Scalp EEG Database, the calculations of the features are performed over a sliding window of one second. The algorithm was evaluated in terms of accuracy, sensitivity, specificity and failure rate. This investigation used SVM as the classification technique. The performance comparisons are made with techniques based on classical features alone, singular value alone and combination of classical features and singular values. The results show that the proposed algorithm achieves better results than using singular values alone or using classical features alone with an average accuracy of 94.82%.

Evaluation of digital speckle filters for ultrasound images

Ultrasound (US) images are inherently corrupted by speckle noise causing inaccuracy of medical diagnosis using this technique. Hence, numerous despeckling filters are used to denoise US images. However most of the despeckling techniques cause blurring to the US images. In this work, four filters namely Lee, Wavelet Linear Minimum Mean Square Error (LMMSE), Speckle-reduction Anisotropic Diffusion (SRAD) and Non-local-means (NLM) filters are evaluated in terms of their ability in noise removal. This is done through calculating four performance metrics Peak Signal to Noise Ratio (PSNR), Ultrasound Despeckling Assessment Index (USDSAI), Normalized Variance and Mean Preservation. The experiments were conducted on three different types of images which is simulated noise images, computer generated image and real US images. The evaluation in terms of PSNR, USDSAI, Normalized Variance and Mean Preservation shows that NLM filter is the best filter in all scenarios considering both speckle noise suppression and image restoration however with quite slow processing time. It may not be the best option of filter if speed is the priority during the image processing. Wavelet LMMSE filter is the next best performing filter after NLM filter with faster speed.

Face recognition technique using Gabor wavelets and Singular Value Decomposition (SVD)

Gabor wavelets (also known as Gabor filters) and Singular Value Decomposition (SVD) have been exploited extensively in the area of face recognition. In this paper, a face recognition system is developed combining features extracted using both Gabor wavelets and SVD. For Gabor wavelets, the extracted feature vectors are selected from only 12 out of 40 Gabor wavelets. The outputs from the 12 filters are selected because it provides relatively more prominent features than the other. This offers the advantage of reducing computational time. As for SVD, only the first five singular values are selected and its associated right singular vectors are used as the feature vectors. The five singular vectors are the one that carry the maximal energy of the image. The combination of Gabor wavelets and SVD offers the advantage of increasing the reliability of the face recognition system. In the face verification stage, the similarity level between facial images is determined by computing the distance between the resulting facial feature vectors obtained from Gabor wavelets and SVD, respectively. The experimental result tested using JAFFE database indicates an average correct acceptance rate of 75.2% and correct rejection rate of 100%. The results show that the combined methods provide a reliable face recognition system.

Speckle reduction of SAR images based on signal subspace technique

In this paper, speckle removal from synthetic aperture radar (SAR) images using subspace-based technique is proposed. The fundamental principle is to decompose the vector space of the noisy image into signal-plus-noise subspace and the noise subspace. Noise reduction is achieved by removing the noise subspace and estimating the clean image from the remaining image subspace. Linear estimation of the clean image is performed by minimizing image distortion while maintaining the residual noise energy below some given threshold. Since the noise is considered to be additive with subspace technique, a homomorphic framework is used to convert the multiplicative speckle noise into additive. The performance of the proposed approach is tested with simulated images and with real SAR images, and compared with Lee filter. The results indicated significant improvements by the proposed technique in terms of structural similarity index measure (SSIM) and equivalent number of looks (ENL).

A subspace-based technique for image noise reduction

In this paper a subspace based technique is proposed for removal of additive white noise in images. The method is based on Time-Domain constrained (TDC) estimator in which the residual noise energy is kept below a threshold while minimizing the signal distortion. The denoising technique employed prewhitening method prior to the subspace filter which is proven to give remarkable results. Experiments were carried out on noise-free images corrupted with simulated additive white noise. The results indicate that the proposed methods provide performance better than Wiener filter in terms of the root mean square error and structural similarity index measure.

A subspace approach for restoring image corrupted by white noise

A new subspace approach is proposed for enhancement of image corrupted by additive white noise. In subspace filtering methods, the noisy image is decomposed into two orthogonal subspaces, a signal subspace and a noise subspace. This decomposition is possible under the assumption of a low-rank model for image and the availability of an estimate of the noise covariance matrix. It is shown in this paper that the proposed image restoration method performs better than the Wiener filtering and wavelet denoising techniques.

2D controlled source electromagnetic modeling in layered medium using hybrid method

This paper presents a hybrid numerical solution of field governing equations of electromagnetic (EM) geophysical method. The proposed method take advantage of implementation simplicity of finite difference (FD) method and the advantage of multiscale resolution of wavelet transform. For evaluation of the proposed method, 2D hypothetical geoelectrical layered-model with simple anisotropy regions are assumed for geophysical modeling. The FD method employed staggered grid scheme, which gives electric field responses on uniform rectangular meshes. Results generated from FD methods give coarse solution of electric field responses. To further refined the electric field value, wavelet expansion is used which produce a better solution that indicates the presence of thin or thick hydrocarbon in the model.