Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
79 results
Search Results
Item Smart visual surveillance technique for better crime management(CRC Press, 2017) Narendra Rao, T.J.; Rajan, J.In the last few years, with concerns about terrorism and global security on the rise, it has become vital to have inplace threat detection devices to detect and recognize suspicious activities and alert the authorities to take appropriate counter actions. Visual Sensor Networks consisting of surveillance cameras are in wide use due to their highly effective monitoring ability which is beyond human capacity. Anomalous event detection is the foremost objective of any automated visual surveillance system. In this chapter, a research work on the development of a smart visual surveillance approach is presented. A detailed discussion of a contextual information based anomalous activity detection and localization algorithm (MSTC) is the main focus of the chapter. It is a modified version of the Spatio-Temporal Compositions (STC) concept introduced earlier. Certain deviations have been incorporated with an objective to reduce time complexity and achieve better performance. In MSTC, the unusual event detection is based on the likelihood of the different regions or ensembles of each query being normal. This requires probabilistic modeling of the arrangement of volumes in the ensembles. The inference mechanisms identify the compositions to be anomalous, which have lower likelihood of occurrence than a threshold. The redundancy of volumes and ensembles has been avoided through their construction in a non-overlapped manner, which aids in lowering the computational complexity of the codebook construction and probabilistic modeling steps. The codebook construction process employed is less complex and has proved to be reliable during experimentation. Three adaptive inference mechanisms determine the decision making threshold for anomalous event detection, given the context. Further, towards making the surveillance approach energy efficient, the concepts of event-driven processing of the query videos and anomaly-driven transmission of only the anomalous frames, implemented in the work have also been included in the chapter. Visual enhancement of the unusual events to aid in surveillance application is another feature discussed. All these software components together demonstrate intelligent functioning in this approach towards smart visual surveillance. © 2018 by Taylor & Francis Group, LLC.Item Anomalous event detection methodologies for surveillance application: An insight(IGI Global, 2017) Rao, T.J.N.; Tahiliani, M.P.; Girish, G.N.; Rajan, J.Automatic visual surveillance systems serve as in-place threat detection devices being able to detect and recognize anomalous activities which otherwise would lead to potentially harmful situations, and alert the concerned authorities to take appropriate counter actions. However, development of an efficient visual surveillance system is quite challenging. Designing an unusual activity detection mechanism which is accurate and real-time is the primary challenge. Review of literature carried out led to the inference that there are some attributes which are essential for a successful unusual event detection mechanism for surveillance application. The desired approach must detect genuine anomalies in real-world scenarios with acceptable accuracy, should adapt to changing environments and, should require less computational time and memory. In this chapter, an attempt has been made to provide an insight into some of the prominent approaches employed by researchers to solve these issues with a hope that it will benefit researchers towards developing a better surveillance system. © 2018 IGI Global. All rights reserved.Item Anomalous event detection methodologies for surveillance application: An insight(IGI Global, 2018) Rao, T.J.N.; Girish, G.N.; Tahiliani, M.P.; Rajan, J.Automatic visual surveillance systems serve as in-place threat detection devices being able to detect and recognize anomalous activities which otherwise would lead to potentially harmful situations, and alert the concerned authorities to take appropriate counter actions. However, development of an efficient visual surveillance system is quite challenging. Designing an unusual activity detection mechanism which is accurate and real-time is the primary challenge. Review of literature carried out led to the inference that there are some attributes which are essential for a successful unusual event detection mechanism for surveillance application. The desired approach must detect genuine anomalies in real-world scenarios with acceptable accuracy, should adapt to changing environments and, should require less computational time and memory. In this chapter, an attempt has been made to provide an insight into some of the prominent approaches employed by researchers to solve these issues with a hope that it will benefit researchers towards developing a better surveillance system. © 2019 by IGI Global.Item Advances in ultrasound despeckling: An overview(IGI Global, 2019) Sudeep, P.V.; Ponnusamy, P.; Rajan, J.The B-mode ultrasound images are corrupted due to the presence of speckle noise. Hence, the speckle removal in the ultrasound images is essential for proper clinical examination and quantitative assessments. The speckle pattern varies with several imaging parameters as well as the anatomical structure in the image. It is hard to avoid speckle by performing averaging and low noise system designs. An excessive speckle reduction diminishes the visibility of small anatomical structures and thereby makes the image understanding complicated. This chapter is intended to encapsulate various techniques for reducing speckle in medical ultrasound images and improving the image quality for visual inspection and/or computer-assisted diagnosis of ultrasound images. In addition, the chapter surveys the papers published between 2015 and 2018 to highlight the latest trends in the despeckling of ultrasound images. The chapter also presents the performance comparison of a few popular algorithms to despeckle medical ultrasound images. © 2019, IGI Global.Item Despeckling algorithms for optical coherence tomography images: A review(IGI Global, 2019) Anoop, B.N.; Girish, G.N.; Sudeep, P.V.; Rajan, J.Optical coherence tomography (OCT) is a non-invasive imaging technique widely used in ophthalmology. The presence of speckle affects the quality of OCT images. Despeckling is necessary to improve its visual quality, and it is an integral part of software packages used for the computerized analysis of OCT. Even though a few methods for despeckling OCT are available in the literature, a cross-comparison of their performance is not known to be available. In this chapter, the techniques available in the literature for despeckling the OCT images have been identified. The results of the despeckling algorithms are compared both qualitatively and quantitatively by concerning the noise suppression capability and feature preservation. Among the available techniques, iterative adaptive unbiased (IAUB) filter is found to be superior as far as its performance regarding despeckling on retinal OCT images. © 2019, IGI Global.Item A Lightweight Convolutional Neural Network Model for Tuberculosis Bacilli Detection From Microscopic Sputum Smear Images(wiley, 2021) Panicker, R.O.; Pawan, S.J.; Rajan, J.; Sabu, M.K.This chapter describes a lightweight convolutional neural network model that automatically detects Tuberculosis (TB) bacilli from sputum smear microscopic images. According to WHO, about onefourth of the population in the universe is infected with TB, and every day five thousand people are killed due to TB disease. There are well-known recommended diagnostics are available for TB detection, among them sputum smear microscopic examination is a primary and most efficient recommended method for most of the developing and moderately developed countries. However, this manual detection method is highly error-prone and time-consuming. In this chapter, we proposed a lightweight CNN model for classifying Tuberculosis bacilli from non-bacilli objects. We adopted a Convolutional Neural Network (CNN) architecture with a skip connection of variable lengths that can identify TB bacilli from sputum smear microscopic images. The performance of the proposed model in terms of accuracy is close to the state-of-the-art. However, the number of parameters in the proposed model is significantly less than other recently proposed models. © 2021 Scrivener Publishing LLC.Item A hybrid model for rician noise reduction in MRI(IEEE Computer Society help@computer.org, 2013) Sudeep, P.V.; Ponnusamy, P.; Rajan, J.Magnetic Resonance Images (MRI) are normally corrupted with random noise mainly arised from the patient's body and from the scanning apparatus. This paper describes a new technique to remove the homogeneous Rician noise in the magnitude magnetic resonance (MR) images. Linear minimum mean square error (LMMSE) estimator is a good choice to solve this inverse problem. In another way, denoising can be considered as a solution for L1 regularization problem of compressed sensing (CS). The Split Bregman iteration technique is effectively used in this stage in order to minimize the total variation (TV) functional. By combining these results in transform domain, the denoising is expected to be improved. Experiments show that the proposed algorithm outperforms other existing methods in the literature in terms of Peak Signal to Noise Ratio (PSNR). © 2013 IEEE.Item A new nonlocal maximum likelihood estimation method for denoising magnetic resonance images(2013) Rajan, J.; den Dekker, A.J.; Juntu, J.; Sijbers, J.Denoising of Magnetic Resonance images is important for proper visual analysis, accurate parameter estimation, and for further preprocessing of these images. Maximum Likelihood (ML) estimation methods were proved to be very effective in denoising Magnetic Resonance (MR) images. Among the ML based methods, the recently proposed Non Local Maximum Likelihood (NLML) approach gained much attention. In the NLML method, the samples for the ML estimation of the true underlying intensity are selected in a non local way based on the intensity similarity of the pixel neighborhoods. This similarity is generally measured using the Euclidean distance. A drawback of this approach is the usage of a fixed sample size for the ML estimation and, as a result, optimal results cannot be achieved because of over- or under-smoothing. In this work, we propose an NLML estimation method for denoising MR images in which the samples are selected in an adaptive way using the Kolmogorov-Smirnov (KS) similarity test. The method has been tested both on simulated and real data, showing its effectiveness. © Springer-Verlag 2013.Item Single image super resolution from compressive samples using two level sparsity based reconstruction(Elsevier B.V., 2015) Nath, A.G.; Nair, M.S.; Rajan, J.Super Resolution based on Compressed Sensing (CS) considers low resolution (LR) image patch as the compressive samples of its high resolution (HR) patch. Compressed sensing based image acquisition systems acquire less number of random linear measurements without first collecting all the pixel values. But using these compressive measurements directly to reconstruct the image causes quality issues. In this paper an image super-resolution method with two level sparsity based reconstruction via patch based image interpolation and dictionary learning is proposed. The first level reconstruction generates a low resolution image from random samples and the interpolation scheme used in this algorithm reduces the HR-LR patch coherency due to neighborhood issue which is a major drawback of single image super resolution algorithms. The dictionary based reconstruction phase generates the high resolution image from the low resolution output of the first level reconstruction phase. The experimental results proved that the proposed two level reconstruction scheme recovers more details of the image and yields improved results from very few samples (around 35-45%) than the state-of-the-art algorithms which uses low resolution image itself as input. The results are compared by considering both PSNR values and visual perception. © 2015 The Authors.Item Coupled PDE for Ultrasound Despeckling Using ENI Classification(Elsevier B.V., 2016) Soorajkumar, R.; Krishnakumar, P.; Girish, D.; Rajan, J.Speckle is a type of noise which is often present in ultrasound images. Speckle is formed due to constructive or destructive interference of ultrasound waves. Due to the granular pattern of speckle noise, it hides important details in ultrasound images. Many despeckling techniques are proposed in the literature, but most of them fail to reach a balance between the removal of speckle noise and preservation of the fine details in the image. In this work, an improved coupled PDE model is proposed which combines second order selective degenerate diffusion (SDD) model and fourth order PDE model based on the assumption that speckle in ultrasound image follows Gamma distribution. An edge noise interior (ENI) method is used to control the diffusion. With the help of ENI controlling function, the diffusion at edge pixels and noisy pixels are selectively accomplished with varying speed. Thus, the proposed model preserves the edges and fine texture details in the image. The model is tested on simulated images after corrupting the images with various levels of Gamma noise. Further, we have tested it on real ultrasound images also. The performance of the proposed model is compared with other similar techniques and the proposed method outperforms other state-of-the-art methods, both in terms of qualitative and quantitative measures. © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license.