Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
78 results
Search Results
Item A novel solution to routing in Cognitive Radio ad-hoc networks in high primary user traffic environments(Institute of Electrical and Electronics Engineers Inc., 2015) Sridhar, A.; Sridhar, S.; Lal, S.Cognitive Radio technology holds great promise in solving the problem of spectrum scarcity. A plethora of routing protocols exist for Cognitive Radio networks, however most of them relay on establishing an end-to-end path using a Common Control Channel. This paper focuses on scenarios where the Primary User traffic is very high and erratic and therefore trying to set up end-to-end paths is not feasible. A novel solution to this problem is proposed where the cognitive users form a Cognitive Delay Tolerant Network through a modification in the network stack. Well researched delay tolerant networking routing protocols designed for networks with unreliable links, configured for multiple channel can used for routing in high primary user traffic environments. Through extensive simulation we show the that proposed architecture provides very high delivery ratio (close to 1) in the presence of very high primary user traffic with negligible computational complexity and the absence of a common control channel. We also show that trying to rely on routing protocols that try to establish end to end paths such as Multi-Channel AODV is not feasible. The performance of Multi-Channel AODV and proposed architecture is compared and analyzed with bundle/packet delivery ratio, end-to-end delay and hop count as performance metrics. © 2015 IEEE.Item Automated bone age assessment using bag of features and random forests(Institute of Electrical and Electronics Engineers Inc., 2018) Simu, S.; Lal, S.The Bone Age is a fairly reliable measure of persons growth and maturation of skeleton. Bone age assessment (BAA) is a procedure that is used to predict the age of a person. The construction of a complete and fairly accurate automated bone age assessment system (ABAA) requires efficient feature extraction and classification methods. In this paper, we have presented an implementation of Bag of Features (BoF) method along with Random Forest classifier on phalanges or bones of fingers. The results have outperformed previous methods available as we have achieved a mean error of 0.58 years and 0.77 years of RMSE for bone age range of 0-18 years. Our experiments have also proved that use of gender bias improves the classification. The best performance was obtained for the ring, middle and index fingers. © 2017 IEEE.Item A Visibility Restoration Framework for rainy images by using Lo gradient minimization and Bilateral Filtering(Institute of Electrical and Electronics Engineers Inc., 2018) Pal, N.S.; Lal, S.; Shinghal, K.In this paper we proposed an algorithm for the visibility restoration of color rainy images. Rain drops affect the visual impact on image so that the visibility and intensity of the captured images are very low. We use a L0 gradient minimization technique for removing the rain pixels. L0 gradient minimization technique globally control the non zero gradient, so that low amplitude and insignificant detail are diminished but salient edges of the images are preserved. First we apply this refining illumination techniques on our rainy image for removing the rain and Secondly we apply a bilateral filtering for smoothing the color image for human perception. The quantitative results of proposed algorithm and other existing rainy algorithms for color rainy images are obtained in terms peak to peak signal to noise ratio (PSNR), universal image quality index(UIQI) and structure similarity index measurement (SSIM) and on different color rainy images from databases. The visual quality of the proposed algorithm also better than other existing rainy algorithms. Finally, the simulation result shows that the proposed algorithm is highly effective and efficient for visibility restoration of rainy images. © 2018 IEEE.Item A Framework for Quality Enhancement of Multispectral Remote Sensing Images(Institute of Electrical and Electronics Engineers Inc., 2018) Suresh, S.; Das, D.; Lal, S.Researches in satellite image enhancement have been particularly confined to two major areas-contrast enhancement and image de noising of remote sensing images. The processing of relatively dark or shadowed images necessitates the need for robust remote sensing enhancement techniques. In this paper, a robust framework for quality enhancement of multispectral remote sensing images is proposed. The quantitative results of proposed algorithm and other existing remote sensing enhancement algorithms are calculated in terms of DE, NIQMC, BIQME, PisDist and CM on different remote sensing and other image databases. Results reveal that visual enhancement of the proposed algorithm is better than other existing remote sensing enhancement algorithms. Finally, the simulation experimental results show that proposed algorithm is effective and efficient for remotes sensing as well as natural images. © 2017 IEEE.Item Performance analysis of despeckling filters for retinal optical coherence tomography images(Institute of Electrical and Electronics Engineers Inc., 2018) Gupta, P.K.; Lal, S.; Husain, F.This paper presents performance analysis of different despeckling filters used for denoising of the optical coherence tomography (OCT) Images. Currently OCT imaging is one of the best technique used in biomedical application to detect the abnormality in the human eye. OCT images normally suffer from granular patterns called speckle noise. Speckle noise is an inherent property of an OCT images which affects the visual quality of the images, hence difficult to diagnosis the patients. Therefore, speckle noise reduction from the OCT images is an important prerequisite, whenever OCT imaging is used for diagnosis. Here, a comparative analysis of different despeckling filters used for the denoising of OCT images is presented. The speckle noise intensity is depends on the various imaging system parameters and on the different structure representations used for the image tissues. A denoising technique is to be designed in such a way that it should be able to reduce the speckle noise from the OCT images while preserve the tissues and fine details of the images. © 2018 IEEE.Item Feature Extraction of Normalized Colorectal Cancer Histopathology Images(Springer Verlag service@springer.de, 2019) Kumar Jain, A.K.; Lal, S.This paper presents different types of feature extraction of normalized colorectal cancer histopathology images. These highlights are exceptionally helpful for separating epithelium and stroma in colorectal cancer (CRC) histopathology images. It is also useful for selecting features and its analysis. In this paper, 27 features are extracted in which 5 are the visual texture features and 22 are the other features such as GLCM, run length and intensity-based features to separate epithelium from the stroma of Colorectal Cancer histopathology images. The utilized component has straightforwardly identified with the human recognition which makes it conceivable to distinguish the nearness of tissue based on parameters. The quantity of utilized highlights is little to differentiate the epithelium from the stroma of CRC histopathology images. In the simulation, we use well-defined and verified histopathology images of stroma and epithelium to correctly differentiate epithelium from stroma. The textural features measure provides the excellent result for 16 typical texture patterns. The issue emerges between the human vision, and modernized strategies that are experienced in this examination show the central point in dissecting of the surface. In which, some of them has removed by using better techniques. In conclusion, perception-based features work well in comparison to previously features used. Some modification like colour normalization of epithelium and stroma image and some of the new features are added because the classification of perception-based features is less. © 2019, Springer Nature Singapore Pte Ltd.Item A robust method for nuclei segmentation of HE stained histopathology images(Institute of Electrical and Electronics Engineers Inc., 2020) Lal, S.; Desouza, R.; Maneesh, M.; Kanfade, A.; Kumar, A.; Perayil, G.; Alabhya, K.; Chanchal, A.K.; Kini, J.Segmentation of histopathology images is an initial and vital step for image understanding. To increase the throughput and to maintain high accuracy, we have to go for an automatic image segmentation method. Here, a robust method for segmentation of cell nuclei in Hematoxylin and Eosin (HE) stained histopathology images is proposed. The proposed segmentation step consists of an initial pre-processing step containing adaptive colour de-convolution and a succession of morphological operations, followed by multilevel thresholding and post-processing steps. Minimum region size is the one parameter which is necessary for this method and set according to the resolution of histopathology image. The proposed nuclei segmentation method does not require any assumptions or prior information about cell morphology. Hence, proposed method applies to the analysis of a wide range of tissues such as liver, kidney, breast, gastric mucosa, and bone marrow and HE stained liver histopathology images from the Hospital. Results yield that proposed nuclei segmentation provides better results in terms of quantitatively and qualitatively on two datasets. © 2020 IEEE.Item DETECTION OF BUILDING INFRASTRUCTURE CHANGES FROM BI-TEMPORAL REMOTE SENSING IMAGES(Institute of Electrical and Electronics Engineers Inc., 2024) Sravya, N.; Kevala, V.D.; Akshaya, P.; Basavaraju, K.S.; Lal, S.; Gupta, D.Change detection (CD) from satellite images is crucial for Earth observation, especially in monitoring urban growth patterns. Recent research has largely focused on using Deep Learning (DL) techniques, particularly variations of Convolutional Neural Network (CNN) architectures. While DL methods have shown promise, many of the models could not preserve the changed areas shape and it fails in predicting the correct edges of changed areas. This paper introduces a CNN based Building Infrastructure Change Detection Network (BICDNet) for predicting changes from bi-temporal remote sensing images. The model leverages a modified Fully Convolutional Siamese-Difference Network to extract detailed features from given images which includes a Multi-Feature Extraction (MFE) block designed to capture features from changed areas of various size within the given input images. To further refine these feature pairs, a modified Atrous Spatial Pyramid Pooling (MASPP) module is integrated, which effectively captures contextual information at multiple scales. The comparison study shows that the proposed BICDNet performs better than the existing CD models. © 2024 IEEE.Item Recent Advances in Urban Expansion Monitoring Through Deep Learning-Based Semantic Change Detection Techniques From Satellite Imagery(Institute of Electrical and Electronics Engineers Inc., 2024) Basavaraju, K.S.; Sravya, N.; Kevala, V.D.; Lal, S.Urban expansion monitoring is essential for understanding and managing the dynamic growth of cities. Recently, deep learning (DL)-based semantic change detection (SCD) techniques have emerged as powerful tools for accurately monitoring urban expansion using satellite imagery. This paper offers comprehensive overview of the recent advancements in urban expansion monitoring through DL-based SCD techniques. It covers various publicly available SCD datasets and assesses performance, advantages, and limitations of existing DL-based SCD architectures, categorized into three types. Furthermore, the paper discusses the challenges encountered in DL-based SCD techniques. Finally, it outlines future research directions in urban expansion monitoring using DL-based SCD techniques. © 2024 IEEE.Item Modified Dual Domain Network for SAR Change Detection(Institute of Electrical and Electronics Engineers Inc., 2024) Kevala, V.D.; Ravi, S.; Surya Kaushik, B.N.; Lal, S.Synthetic Aperture Radar (SAR) images are utilised for change detection analysis due to their all-weather imaging capabilities. This paper proposes modified dual domain network (MDDNet) for SAR change detection. We introduced the atrous spatial pyramid pooling block to extract multiscale characteristics in the spatial domain. The MDDNet extracts features from both the spatial and frequency domains. The proposed network is trained unsupervised with pre-classification output. The performances of proposed and existing SAR change detection models are evaluated on four bitemporal SAR datasets. The experimental results indicate that the results of proposed MDDNet is better than existing change detection models on four bitemporal SAR dataset. © 2024 IEEE.