Ravindra Dhuli

Fusion of Infrared and Visible Sensor Images Based on Anisotropic Diffusion and Karhunen-Loeve Transform

Image fusion is a process of generating a more informative image from a set of source images. Major applications of image fusion are in navigation and military. Here, infrared and visible sensors are used to capture complementary images of the targeted scene. The complementary information of these source images has to be integrated into a single image using some fusion algorithms. The aim of any fusion method is to transfer maximum information from the source images to the fused image with a minimum information loss. It has to minimize the artifacts in the fused image. In this paper, we propose a new edge preserving image fusion method for infrared and visible sensor images. Anisotropic diffusion is used to decompose the source images into approximation and detail layers. Final detail and approximation layers are calculated with the help of Karhunen-Loeve transform and weighted linear superposition, respectively. A fused image is generated from the linear combination of final detail and approximation layers. Performance of the proposed algorithm is assessed with the help of petrovic metrics. The results of the proposed algorithm are compared with the traditional and recent image fusion algorithms. Results reveal that the proposed method outperforms the existing methods.

Classification of imbalanced ECG beats using re-sampling techniques and AdaBoost ensemble classifier

Abstract Computer-aided heartbeat classification has a significant role in the diagnosis of cardiac dysfunction. Electrocardiogram (ECG) provides vital information about the heartbeats. In this work, we propose a method for classifying five groups of heartbeats recommended by AAMI standard EC57:1998. Considering the nature of ECG signal, we employed a non-stationary and nonlinear decomposition technique termed as improved complete ensemble empirical mode decomposition (ICEEMD). Later, higher order statistics and sample entropy measures are computed from the intrinsic mode functions (IMFs) obtained from ICEEMD on each ECG segment. Furthermore, three data level pre-processing techniques are performed on the extracted feature set, to balance the distribution of heartbeat classes. Finally, these features fed to AdaBoost ensemble classifier for discriminating the heartbeats. Simulation results show that the proposed method provides a better solution to the class imbalance problem in heartbeat classification.

Fusion of MRI and CT images using guided image filter and image statistics

In medical imaging using different modalities such as MRI and CT, complementary information of a targeted organ will be captured. All the necessary information from these two modalities has to be integrated into a single image for better diagnosis and treatment of a patient. Image fusion is a process of combining useful or complementary information from multiple images into a single image. In this article, we present a new weighted average fusion algorithm to fuse MRI and CT images of a brain based on guided image filter and the image statistics. The proposed algorithm is as follows: detail layers are extracted from each source image by using guided image filter. Weights corresponding to each source image are calculated from the detail layers with help of image statistics. Then a weighted average fusion strategy is implemented to integrate source image information into a single image. Fusion performance is assessed both qualitatively and quantitatively. Proposed method is compared with the traditional and recent image fusion methods. Results showed that our algorithm yields superior performance.

Classification of ECG heartbeats using nonlinear decomposition methods and support vector machine

Classifying electrocardiogram (ECG) heartbeats for arrhythmic risk prediction is a challenging task due to minute variations in the amplitude, duration and morphology of the ECG signal. In this paper, we propose two feature extraction approaches to classify five types of heartbeats: normal, premature ventricular contraction, atrial premature contraction, left bundle branch block and right bundle branch block. In the first approach, ECG beats are decomposed into intrinsic mode functions (IMFs) using ensemble empirical mode decomposition (EEMD). Later four parameters, namely the sample entropy, coefficient of variation, singular values, and band power of IMFs are extracted as features. In the second approach, the same features are computed from IMFs extracted using an empirical mode decomposition (EMD) algorithm. The features obtained from the two approaches are independently fed to the sequential minimal optimization-support vector machine (SMO-SVM) for classification. We used two arrhythmia databases for our evaluation: MIT-BIH and INCART. We compare the proposed approaches with existing methods using the performance measures given by the average values of (i) specificity, (ii) sensitivity, and (iii) accuracy. The first approach demonstrates significant performance with 98.01% sensitivity, 99.49% specificity, and 99.20% accuracy for the MIT-BIH database and 95.15% sensitivity, 98.37% specificity, and 97.57% accuracy for the INCART database.

Super-Resolution Image Reconstruction Using Dual-Mode Complex Diffusion-Based Shock Filter and Singular Value Decomposition

Super-resolution (SR) algorithms are widely used to overcome the hardware limitations of the low-cost image acquisition devices. In this paper, we present a single image SR (SISR) approach in wavelet domain, which simultaneously preserves the contrast and edge information. Our algorithm uses the notion of geometric duality to generate the initial estimation of unknown high-resolution (HR) image, by applying covariance-based interpolation. State-of-the-art wavelet techniques for SISR provide resolution enhancement by replacing the low-frequency subband with the input low-resolution image. This leads to non-uniform illumination in the super-resolved image. The proposed method exploits singular value decomposition to correct the low-frequency subband, as obtained via stationary wavelet transform (SWT). The modified low-frequency subband and the high-frequency subband images are subjected to Lanczos interpolation. The interpolated subbands are filtered by employing diffusion-based shock filter, which operates in the dual dominant modes. All the filtered subband images are fused to generate the final HR image, by applying inverse SWT. Our experimental analysis has demonstrated the superiority of the proposed method in preserving the edges with uniform illumination.

Image super-resolution using dictionaries and self-similarity

Image super resolution attempts to extract a high resolution image using one or more corrupted low resolution images. Typical sparse dictionary based super resolution methods remove the undesired effects but may not significantly enhance resolution. In contrast, methods that exploit local self-similarity enhance the native resolution as well as the undesired artifacts present in the low resolution image. In this paper, we propose a novel single image super resolution approach that renders high resolution images by exploiting dictionary based non-local methods and uses local similarity of small spatial patches of the image to eliminate undesired artifacts. Our quantitative results on several test datasets are promising.

Shock filter based image super-resolution using dual-tree complex wavelet transform and singular value decomposition

Purpose – The purpose of this paper is to present an improved wavelet-based approach in single image super resolution (SISR). The proposed method generates high resolution (HR) images by preserving the image contrast and edges simultaneously. Design/methodology/approach – Covariance-based interpolation algorithm is employed to obtain an initial estimate of the unknown HR image. The proposed method preserves the image contrast, by applying singular value decomposition (SVD)-based correction on the dual-tree complex wavelet transform coefficients. In addition, the dual operating mode diffusion based shock filter (DBSF) ensures noise mitigation and edge preservation. Findings – Experimental results on various test images reveal superiority of the proposed method over the existing SISR techniques in terms of peak signal to noise ratio, structural similarity index measure and visual quality. Originality/value – With SVD-based correction, the proposed method preserves the image contrast and also the DBSF operat...

Edge preserving super-resolution algorithm using multi-stage cascaded joint bilateral filter

Super-resolution (SR) algorithms address the inabilities of poor imaging devices, there by producing high quality images with enhanced resolution. We propose a new SR approach which produces sharp high resolution (HR) image using its low resolution (LR) counterparts. The proposed method uses geometric duality for spatially adapting covariance-based interpolation (CBI). To preserve edge information, a multi-stage cascaded joint bilateral filter (MSCJBF) is proposed as an intermediary stage. These edges are incorporated in the high frequency subbands obtained by the stationary wavelet transform (SWT), through nearest neighbor interpolation (NNI) method. Prior to the NNI process, the high frequency subbands undergo two-lobed lanczos interpolation to achieve the desired resolution enhancement. The quantitative and qualitative analysis for various test images prove the superiority of our method.

Low power and low complexity implementation of LPTV interpolation filter

This paper presents an architecture for low power and low complexity implementation of a linear periodically time varying (LPTV) interpolation filter using thread decomposition (TD) technique which decomposes a filter into finite computational threads. TD technique enables us to develop the proposed architecture as a generalization to linear time invariant (LTI) filter structure. The area complexity of the proposed architecture is significantly reduced by optimizing the concurrent threads of the conventional design. Reduction of power consumption is achieved in the proposed design by eliminating futile multiplications and reducing the operating frequency of the multipliers. It involves nearly one fourth the number of adders, multipliers and delay elements compared to the conventional design. The proposed structure is implemented on Virtex FPGA 2vp30-7ff896. From the synthesis results, it is found that the proposed design offers 35.7% reduction in power consumption and 20.6% reduction in device utilization over the conventional design.

Design and analysis of matrix Wiener synthesis filter for multirate filter bank

Abstract In this paper, we present a solution to the problem of reconstructing the input of a maximally decimated filter bank from the subband components using Wiener filtering. We present a generalized structure for applying Wiener filtering at the output of the analysis stage of a uniform filter bank (UFB). This structure can be used to model a situation where the desired signal is a filtered version of the input signal. Some interesting results for matrix inversion are derived and used to reduce the complexity of the Wiener filter expression. The resulting expression provides many insights into the properties of the Wiener synthesis filter designed. The Wiener synthesis filter turns out to be independent of the input spectral properties. The proposed Wiener synthesis filter bank exploits the pseudocirculant property. Thus all distortions are completely removed and the filter bank reduces to a linear time invariant (LTI) filter of interest. We later extend the analysis to non-uniform filter banks (NUFBs).