Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
17 results
Search Results
Item A Class of Frozen Regularized Gauss-Newton Methods Under Weak Conditions(Springer, 2025) George, S.; Jidesh, P.Qinian Jin (2010) studied Frozen Regularized Gauss Newton Method (FRGNM) for approximating a solution of nonlinear ill-posed equation. The assumptions used to prove results in Jin’s paper are too restrictive. In this study, we analyze the convergence of FRGNM under weaker assumptions. This way we extend the applicability of FRGNM to the problems which does not satisfy the assumptions in Qinian Jin (2010). We also provide numerical results obtained for five different parameter choice strategies for FRCNM. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.Item Steganalysis: Using the blind deconvolution to retrieve the hidden data(2011) Jidesh, P.; George, S.Steganography has gained a substantial attention due to its application in wide areas. Steganography as it literally mean is hiding the information (stego data) inside the data (communication data) so that the receiver can only extract the desired information from the data. Steganalysis is the reverse process of steganography in which the information about the original data is hardly available, from the received data the extractor needs to identify the original data. Since this belong to a class of inverse problems it is hard to find the approximate match of the original data from the received one. In most of the cases this will fall under the category of ill-posed problems. The stego-data that has been embedded into the communication data can be considered as linear bounded operator operating on the input data and the reverse process (the Steganalysis) can be thought like a deconvolution problem by which we can extract the original data. Here we are assuming the watermarking as a linear operation with a bounded linear operator K : X→Y where X and Y are spaces of Bounded Variation (BV). The forward problem (the Steganography) is a direct convolution and the reverse (backward) problem (steganalysis) is a de-convolution procedure. In this work we are embedding a Gaussian random variable array with zero mean and with a specific variance into the data and we show how the original data can be extracted using the regularization method. The results are shown to substantiate the ability of the method to perform steganalysis. © 2011 IEEE.Item Shock coupled coherence enhancing diffusion for robust core-point detection in fingerprints(2011) Jidesh, P.; George, S.Enhancing the flow-like structures is important in forensic applications especially in fingerprint analysis. In most of the practical scenarios the poor quality of the off-line prints collected, adversely affect the verification process. Though there has been a plethora of methods proposed in literature for enhancing the degraded images, very few of them are suitable for enhancing the flow-like structures because they are ignorant of the coherence features present in images with dominant flow-like structures. In this paper we propose a method which enhances the fingerprint images with utmost consideration to the coherence features of the fingerprints. This method provides a shock at the inflection points while retaining the flow-like nature of the fingerprints. In other words, it enhances the coherence of features along with the edges. The experimental results shown endorses on the capability of the method to enhance the fingerprints which in turn will result in identification of the core-points in the fingerprints with a better accuracy. The core-point identification is a crucial step in fingerprint verification. © 2011 IEEE.Item An adaptive total variation model with local constraints for denoising partially textured images(2011) Bini, A.A.; Bhat, M.S.; Jidesh, P.Denoising algorithms such as Total Variation model modify smooth areas in images into piecewise constant patches and small scale details and textures present in the original image are not preserved satisfactorily by these processes. In this paper, we present an algorithm based on an adaptive Total Variation norm of the gradient of the image, with a family of local constraints for efficient denoising of natural images. In fact, natural images consist of smooth and textured regions. Staircase effect is reduced in smooth areas by using a modified Total Variation functional. The set of local constraints, one for each pixel in the image are able to preserve most of the fine details and textures in the images. Visual and quantitative results of proposed method are presented and are compared with results of existing methods. © 2011 Copyright Society of Photo-Optical Instrumentation Engineers (SPIE).Item Quantification and morphology studies of nanoporous alumina membranes: A new algorithm for digital image processing(2013) Choudhari, K.S.; Jidesh, P.; Sudheendra, P.; Kulkarni, S.D.A new mathematical algorithm is reported for the accurate and efficient analysis of pore properties of nanoporous anodic alumina (NAA) membranes using scanning electron microscope (SEM) images. NAA membranes of the desired pore size were fabricated using a two-step anodic oxidation process. Surface morphology of the NAA membranes with different pore properties was studied using SEM images along with computerized image processing and analysis. The main objective was to analyze the SEM images of NAA membranes quantitatively, systematically, and quickly. The method uses a regularized shock filter for contrast enhancement, mathematical morphological operators, and a segmentation process for efficient determination of pore properties. The algorithm is executed using MATLAB, which generates a statistical report on the morphology of NAA membrane surfaces and performs accurate quantification of the parameters such as average pore-size distribution, porous area fraction, and average interpore distances. A good comparison between the pore property measurements was obtained using our algorithm and ImageJ software. This algorithm, with little manual intervention, is useful for optimizing the experimental process parameters during the fabrication of such nanostructures. Further, the algorithm is capable of analyzing SEM images of similar or asymmetrically porous nanostructures where sample and background have distinguishable contrast. Copyright © Microscopy Society of America 2013.Item Geometric transform invariant Brain-MR image analysis for tumor detection(2013) Tom, A.; Jidesh, P.In this work we propose a translational, rotational and scaling invariant scheme for possible detection of tumors in Brain-Magnetic Resonance (MR) images. The method incorporates the features like shape, position and texture to accurately diagnose from the infected images. The geometric transformation invariant nature of the method helps in detecting the tumor in various scales, positions and orientations, at a better rate compared to the state-of-the art methods. The method combines three features (shape, position and texture) to form a feature vector, which is used for detecting the infected parts in the image. In order to improve the accuracy of detection process, we employ a preprocessing step to denoise and enhance the images. The result section details the analysis and results of the proposed method and highlights on the accuracy of the method to properly identify the tumor parts in an MR image. © 2013 IEEE.Item A complex diffusion driven approach for removing data-dependent multiplicative noise(2013) Jidesh, P.; Bini, A.A.In this paper we propose a second-order non-linear PDE based on the complex diffusion function. The proposed method exhibits better restoration capability of ramp edges in comparison to other second-order methods discussed in the literature. The proposed model is designed for Gamma distributed multiplicative noise which commonly appears in Ultra Sound (US) and Synthetic Aperture Radar (SAR) images. The fidelity/reactive term augmented to the complex diffusive term is derived based on the Bayesian maximum a posteriori probability (MAP) estimator as detailed in Aubert and Ajol ([10]). The regularization parameter is selected based on the noise variance of the image and thus this adaptive method helps in restoring the images at various noise variances without manually fixing the parameter. The results shown in terms of both visual and qualitative measures demonstrate the capability of the model to restore images from their degraded observations. © Springer-Verlag 2013.Item Non-local Gradient Fidelity Model for Multiplicative Gamma Noise Removal(Institute of Electrical and Electronics Engineers Inc., 2018) Banothu, B.; Jidesh, P.In this paper a non-local gradient vector flow model is designed for restoration of images corrupted with Gamma distributed (speckle) noise and linear blurring artefacts. The filter effectively preserves edges and finer details in the course of its evolution due to the presence of the non-local TV based diffusion term and the piecewise linear approximation is reduced considerably by the gradient fidelity term present in the model. The model is found suitable for restoration of various images from the field of satellite and clinical imaging. The experimental results are shown and compared for different image data sets both visually and qualitatively using various statistical measures. © 2017 IEEE.Item Lung nodule identification and classification from distorted CT images for diagnosis and detection of lung cancer(Springer Verlag service@springer.de, 2019) Savitha, G.; Jidesh, P.An automated computer-aided detection (CAD) system is being proposed for identification of lung nodules present in computed tomography (CT) images. This system is capable of identifying the region of interest (ROI) and extracting the features from the ROI. Feature vectors are generated from the gray-level covariance matrix using the statistical properties of the matrix. The relevant features are identified by adopting principle component analysis algorithm on the feature space (the space formed from the feature vectors). Support vector machine and fuzzy C-means algorithms are used for classifying nodules. Annotated images are used to validate the results. Efficiency and reliability of the system are evaluated visually and numerically using relevant measures. Developed CAD system is found to identify nodules with high accuracy. © Springer Nature Singapore Pte Ltd 2019.Item Retinal Vessel Classification Using the Non-local Retinex Method(Springer, 2020) Smitha, A.; Jidesh, P.; Febin, I.P.Automatic retinal vessel segmentation has turned out to be highly propitious for medical practitioners to diagnose diseases like glaucoma and diabetic retinopathy. These diseases are classified based on the thickness of the retinal vessel, the pressure imposed on the nerve endings and optical disc to cup ratio of the retina. The state-of-the-art device for this purpose presently available in the market is expensive and has scope to meliorate sensitivity and precision of its performance. Thus, automatic retinal blood vessel segmentation and classification is the need of the hour. In this paper, a novel non-local total variational retinex based retinal image preprocessing approach is proposed to extract the retinal vessel features and classify the vessels using ground truth images. Matlab implementation results indicate that an average accuracy of 94% with an acceptable range of sensitivity and specificity could be achieved on the retinal image database available online. © 2020, Springer Nature Switzerland AG.