R. S. Anand

A simple design method for the cosine-modulated filter banks using weighted constrained least square technique

Abstract This paper presents a simple and efficient design method for cosine-modulated filter banks with prescribed stopband attenuation, passband ripple, and channel overlap. The method casts the design problem as a linear minimization of filter coefficients such that their value at ω = π /2 M is 0.707, which results in a simpler, more direct design procedure. The weighted constrained least squares technique is exploited for designing the prototype filter for cosine modulation (CM) filter banks. Several design examples are included to show the increased efficiency and flexibility of the proposed method over the exiting methods. An application of the proposed method is considered in the area of sub-band coding of the ECG and speech signals.

An Improved Method for Designing Quadrature Mirror Filter Banks via Unconstrained Optimization

This paper proposes an algorithm to design a two-channel linear phase quadrature mirror filter (QMF) bank. The design problem is presented systematically as an unconstrained optimization that minimizes the weighted sum of error of transfer function of the filter bank at quadrature frequency, stopband energy and the passband error of a prototype filter (PF). A new method is developed for the design of a low pass prototype filter for QMF banks. For solving given optimization problem, Quasi-Newton optimization technique is used. Numerical examples and comparisons with several existing methods are included to show the performances and effectiveness of this method. An application of the proposed method is considered in the area of subband coding of the images.

An improved method for the design of quadrature mirror filter banks using the Levenberg–Marquardt optimization

This paper presents an improved and efficient method for the design of a two-channel quadrature mirror filter (QMF) bank. In the proposed method, the filter bank design problem is formulated as a low-pass prototype filter design problem, whose responses in the passband and stopband are ideal and their filter coefficients value at quadrature frequency is 0.707. A new method is developed for the design of a low-pass prototype filter which minimizes the objective function by optimizing the filter taps weights using the Levenberg–Marquardt method. When compared with other existing algorithms, it significantly reduces peak reconstruction error (PRE), error in passband, stopband and transition band. Several design examples are included to show the increased efficiency and the flexibility of the proposed method over existing methods. An application of this method is considered in the area of subband coding of the ultrasound images.

An improved closed form design method for the cosine modulated filter banks using windowing technique

This paper presents an improved closed form method for designing cosine modulated filter banks with prescribed stopband attenuation and channel overlap. The proposed method is based on optimum passband frequency, which is calculated using empirical formulas instead of using time consuming single or multivariable optimization. Different window functions are employed to design the prototype filter with the novelty of exploiting spline functions in the transition region of the ideal filter. Several design examples are included to show the increased efficiency of the proposed method over other exiting methods in terms of computation time, amplitude distortion (eam) and aliasing distortion (ea). An application of the proposed method is considered in the area of subband coding of the ECG and speech signals.

Multiresolution local binary pattern variants based texture feature extraction techniques for efficient classification of microscopic images of hardwood species

Multiresolution local binary pattern (MRLBP) variants based texture feature extraction technique. A multiresolution local binary pattern (MRLBP) variants based texture feature extraction techniques for hardwood species categorization into 75 classes.Feature dimension reduction using principal component analysis.Investigation of the effectiveness of MRLBP variants texture features for hardwood species classification using LDA, linear SVM and RBF kernel SVM classifiers. In this paper, multiresolution local binary pattern (MRLBP) variants based texture feature extraction techniques have been proposed to categorize hardwood species into its various classes. Initially, discrete wavelet transform (DWT) has been used to decompose each image up to 7 levels using Daubechies wavelet (db2) as decomposition filter. Subsequently, six texture feature extraction techniques (local binary pattern and its variants) are employed to obtain substantial features of these images at different levels. Three classifiers, namely, linear discriminant analysis (LDA), linear and radial basis function (RBF) kernel support vector machine (SVM), have been used to classify the images of hardwood species. Thereafter, classification results obtained from conventional and MRLBP variants based texture feature extraction techniques with different classifiers have been compared. For 10-fold cross validation approach, texture features acquired using discrete wavelet transform based uniform completed local binary pattern (DWTCLBPu2) feature extraction technique has produced best classification accuracy of 97.40?1.06% with linear SVM classifier. This classification accuracy has been achieved at the 3rd level of image decomposition using full feature (1416) dataset. Further, reduction in dimension of texture features (325 features) by principal component analysis (PCA) has been done and the best classification accuracy of 97.87?0.82% for DWTCLBPu2 at the 3rd level of image decomposition has been obtained using LDA classifier. The DWTCLBPu2 texture features have also established superiority among the MRLBP techniques with reduced dimension features for randomly divided database into fix training and testing ratios.

Context based medical image compression with application to ultrasound images

The basic goal of medical image compression is to reduce the bit rate and enhance the compression efficiency for the transmission and storage of the medical imagery while maintaining an acceptable diagnostic image quality. Because of the storage, transmission bandwidth and the limitations of the conventional compression methods, the medical imagery need to be compressed selectively to reduce the transmission time and storage cost along with the preservance of the high quality of the image. The other important reason of context based medical image compression is the high spatial resolution and contrast sensitivity requirements. In medical images, contextual region is an area which contains the most useful and important information and must be coded carefully without appreciable distortion. A novel scheme for context based coding is proposed here and yields significantly better compression rates than the general methods of JPEG and JPEG2000. In the proposed method the contextual part of the image is encoded selectively on the high priority basis with a very low compression rate (high bpp) and the background of the image is separately encoded with a low priority and a high compression rate (low bpp). As a result, high over all compression rates, better diagnostic image quality and improved performance parameters (CR, MSE, PSNR and CoC) are obtained. The experimental results have been compared to the Scaling, Maxshift, Implicit and EBCOT methods on ultrasound medical images and it is found that the proposed algorithm gives better and improved results.

A simple method for the design of cosine-modulated pseudo-QMF banks

In this paper, a simple and efficient iterative technique for designing of Cosine-Modulated (CM) filter banks with approximate reconstruction is presented. In the proposed technique, cut-off frequency is optimised to satisfy a new Perfect Reconstruction (PR) condition at frequency (ω = π/2M). Different window functions have been used to design the prototype filter for CM filter banks. The simulation results illustrate the improvement achieved through this method over other existing methods in terms of amplitude distortion (eam), Number of Iterations (NOI), aliasing distortion (ea) and computation time (CPU time). When it is exploited for subband coding of the Electrocardiogram (ECG) and speech signals, the proposed method yields good fidelity performance measuring parameters.

An Improved Method for Designing Prototype Filter for M-Band Pseudo QMF Banks

This paper presents an iterative algorithm for the design of multichannel cosine modulated pseudo quadrature mirror filter (QMF) banks. In this algorithm, passband frequency is optimized to satisfy new perfect reconstruction condition at frequency ( ). The proposed algorithm is simple, and has high design efficiency. The performance of this algorithm is evaluated in terms of computational time (CPU time), reconstruction error (RE), number of iterations (NOI), aliasing distortion (AD) and signal to noise ratio (SNR). Several design examples are included to illustrate the proposed algorithm and its improved performances over other exiting methods.

A simple iterative technique for the design of cosine modulated pseudo QMF banks

This paper presents a simple and efficient iterative technique for the design of cosine modulated pseudo quadrature mirror filter (QMF) banks. In the proposed algorithm, cutoff frequency is optimized to satisfy a new perfect reconstruction condition at frequency (ω = π/2M). Adjustable Window functions are used to design the prototype filter cosine modulated pseudo QMF banks. When compared to other existing techniques, the proposed algorithm significantly reduces amplitude distortion, number of iterations, aliasing distortion, and computation time. Several design examples are included to illustrate the proposed algorithm and its improved performances over other exiting methods.

Binary wavelet transform-based completed local binary pattern texture descriptors for classification of microscopic images of hardwood species

This paper presents an approach for generating a binary wavelet transform-based completed local binary pattern (BWTCLBP) texture descriptor to improve the classification accuracy of microscopic images of hardwood species. Firstly, gray-level slicing method is used to obtain eight (b0–b7) bit planes from grayscale image. Then, the two-dimensional binary wavelet transform (2D-BWT) decomposes each of the most significant bit-plane (b7) images up to the fifth scale of decomposition. The texture descriptors are then acquired from each of the subimages up to the five scales of decomposition. Further, two variants of support vector machine (SVM), linear SVM and radial basis function kernel SVM, were employed as classifiers. The classification accuracy of the proposed and existing texture descriptors was compared. The BWT-based uniform completed local binary pattern (BWTCLBPu2) texture descriptor achieved the best classification accuracy of 95.07xa0±xa00.72% at the third scale of decomposition. The classification accuracy is produced by linear SVM classifier for full feature (1416) vector data. In order to overcome the effect of curse of dimensionality, the minimal-redundancy–maximal-relevance feature selection method is employed to select the best subset of feature vector data. This approach has resulted in improved classification accuracy of 96.60xa0±xa00.80% (450) by linear SVM classifier.