Earticle

The proposed algorithm for gene prediction compares the N/3 spectral components of DNA signal with the corresponding spectrum of period-3 suppressed DNA signal. In the DNA N/3 spectrum, for the bases for which the difference between these two spectrums is within a predefined threshold level, the signal values are replaced by the difference signal of the two spectrums. This substitution suppresses the noise in the non-coding regions of N/3 spectrum, while the coding region signal values are not affected, resulting in an improved detection. Performance of this algorithm when evaluated on HMR 195 dataset, in terms of area under the ROC curve, exhibits a 5% improvement on the DFT-based spectral content measure.

Biometric methods have been played important roles in personal recognition during last twenty years. These methods include the face recognition, finger print and iris recognition. Recently iris imaging has many applications in security systems. The aim of this paper is to design and implement a new iris recognition algorithm. In this paper, the new feature extraction methods according to ridgelet transform and curvelet transform for identifying the iris images are provided. At first, after segmentation and normalization the collarette area of iris images has been extracted. Then we improve the quality of image by using median filter, histogram equalization, and the two-dimensional (2D) Wiener filter as well. Finally the ridgelet transform and curvelet transform are applied for extracting features and then the binary bit stream vectors are generated. The Hamming distance (HD) between the input bit stream vector and stored vectors is calculated for iris identification. The experimental results show efficiency of our proposed method.

A 2D-video processing technique for automatic detection of incoming vehicles at night light conditions is a challenging task for any Advanced Driver Assistance Systems (ADAS). We present a novel image processing technique to be used by ADAS to detect and track incoming vehicle’s front headlamp for estimating real time velocity under such conditions. To capture the incoming traffic a standard C-MOS camera is mounted in the front panel of the vehicle and pre-processing filters are applied on each frame for removing impulse noise. Headlamp blobs are segmented from the rectified frame using intensity thresholds. Area of the blob is calculated based on the motion data and is used to predict the real time velocity of the vehicle. Optical flow concept is applied to predict Motion in the headlamps and tracked by kalman filtering. Experiments are conducted to estimate the training set parameters of the distance function. Results that demonstrate system’s high velocity prediction rates using the best practices of image processing and optical flow are presented.

The fractional Fourier transform (FRFT), which is considered as a generalization of the Fourier transform (FT), has emerged as a very efficient mathematical tool in signal processing for signals which are having time-dependent frequency component. The FRFT has an advantage over other transforms being used in the application areas like: signal processing and optics. Some of the properties are still not established or defined in FRFT domain. An effort is made to derive the correlation theorems for FRFT along with the establishment of their respective properties. The proposed auto-correlation theorem is also used to determine the power spectral density of frequency modulated (FM) signal. The results are found in conformity with the standard one.

All the time, there is a demand of High-Resolution (HR) images in electronic imaging applications. Super-Resolution (SR) is an approach used to restore High-Resolution (HR) image from one or more Low-Resolution (LR) images. The goal of SR is to extract the independent information from each LR image in that set and combine the information into a single high resolution (HR) image. The quality of reconstructed SR image obtained from a set of LR images depends upon the registration accuracy of LR images. In this paper SR reconstruction using a sub-pixel shift image registration and Fast Discrete Curvelet transform (FDCT) for image interpolation is proposed. The Curvelet transform is a multiscale pyramid with many directions and positions at each scale. Image interpolation is performed at the finest level in Curvelet domain. Experimentation based results have shown appropriate improvements in PSNR and MSE. Also, it is experimentally verified that the computational complexity of the SR algorithm is reduced.

Many vision tasks are based on edge/corner detection. We propose some rules and utilize them to find suitable window for edge/corner detection by using quadratic programming. Experimental results over synthetic images and real images show satisfactory outcomes.

A technique for integral image compression is presented. The proposed technique relies on applying principle component analysis, PCA, on the wavelet coefficients of the elemental images to improve the quality of the recovered 3D image while achieving high compression ratio. The wavelet coefficients of the individual elemental images are stacked and rearranged before applying PCA compression. The PCA compression is applied to each sub-band individually to enhance the compression ratio. The quality of the reconstructed 3D images and received elemental images are calculated. Results show high compression ratio compared to PCA alone compression while maintaining the recovered 3D image quality. PSNR is used to measure the reconstructed 3D image quality.

Connected digits recognition is important in many applications such as voice-dialing telephone, automated banking system, automatic data entry, PIN entry, etc. In this paper robust features such as Revised perceptual linear prediction (RPLP), Bark frequency cepstral coefficients (BFCC) and Mel frequency perceptual linear prediction (MF-PLP) are used for speaker-independent connected Hindi digits recognition in clean and noisy environments. The recognition performance of these features is compared with recognition performance of Mel frequency cepstral coefficient (MFCC), ΔMFCC and Perceptual linear prediction (PLP) features. MF-PLP features have shown best recognition efficiency for clean as well as for noisy database. MFCC features are calculated by using feature extraction tool of Hidden Markov model Toolkit (HTK). All other features except MFCC are calculated using Matlab and saved in HTK format. Training and testing for speech recognition is done using HTK.

A hybrid denoising method is presented as a combination of Empirical Mode Decomposition (EMD) and Higher Order Statistics (HOS). EMD, an adaptive data-driven method, is used for effective decomposition of a noisy signal into its functional components. Then Kurtosis and Bispectrum operate as Gaussianity estimators, supplemented by Bootstrap techniques, ensuring detection and removal of the signal’s Gaussian components. Thresholding techniques are used at the final step for maximum suppression of signal noise, where thresholds are set by estimating the long-term correlation of the corrupting colored noise in the form of the Hurst exponent. Experimental results prove the applicability of the method in signal denoising. Specifically, EMD-HOS outperforms similar denoising techniques based on Wavelets for the most types of test signals.

Most of the sub-pixel edge detection methods proposed in literature are based on Ghosal and Mehrotra’s method which uses Zernike moments. Some research has been reported using Fourier-Mellin moments for sub-pixel edge detection. Pseudo Zernike moments have been proved to be superior to Zernike moments in terms of their feature representation capabilities and sensitivity to image noise. This paper proposes the use of Pseudo Zernike moments for sub-pixel edge detection. Testing of the proposed algorithm shows significantly better results than Zernike and Fourier-Mellin moment based methods in case of images corrupted with additative or multiplicative noise .

A new approach to modified S-transform based non-stationary power signal de-noising is presented in this paper. Modified S-transform is employed in de-noising techniques to separate high frequency noise which is prevalent in practical signals like power disturbance signals from the low frequency undistorted signals. In this we have tried to bring out the advantage of the modified S-transform and its application in power signal analysis through various examples. S-transform with modified Gaussian window is found to provide excellent normalized frequency contours of the power signal disturbances suitable for accurate detection, localization and classification. The implementation of the de-noising scheme has been carried out in FPGA using non-stationary power signals.