Sumathi Mahadevan
Sumathi Mahadevan has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- S. N. Sivanandam
Computer Science and Engineering Department, Karpagam College of Engineering, Coimbatore, India-641 032
3 shared publications - Mahalakshmi Priya Rajendran
Research Scholar, Madurai Kamaraj University (MKU), Madurai, Tamilnadu, India
2 shared publications S. Kumar Chandar
Christ University, Bangalore - 560029, Karnataka
2 shared publicationsC. Sulochana
Madurai Kamaraj University,Department of Computer Science,Madurai,Tamilnadu,India
1 shared publication- G. B. GovindaPrabhu
Research Scholar, Madurai Kamaraj University (MKU), Madurai, Tamilnadu, India
1 shared publication - Kumar Chandar Sivalingam
Christ University
1 shared publication
Latest Publications
- 2 April 2025
- Published by Springer Nature in Iran Journal of Computer Science
- Vol. 8 (3) , 963-988
PollenMorph AI: quantum contours based segmentation and deep learning for pollen recognition using microscopic images
- 2 April 2025
- Published by Springer Nature in Iran Journal of Computer Science
- Vol. 8 (3) , 963-988
PollenMorph AI: quantum contours based segmentation and deep learning for pollen recognition using microscopic images
- 17 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
An Integrated Approach to Bacteria Structure Detection using Frangi-Thresholding Segmentation and its Impact on Analysis
Molecular characterization, classification, and other analytical tasks in microbiology require accurate segmentation of bacteria structures in microscopic images. Although bacterial shapes and textures are heterogeneous, reliable segmentation remains challenging because of low contrast, illumination variations, and variations in illumination. From microscopy scans, a pipeline for integrating image processing is presented in this work. To enhance image quality and focus on diagnostically relevant areas, an adaptive region-of-interest cropping approach is combined with Retinex enhancement, color balancing, and bi-histogram equalization. Felzenszwalb's graph-based active contour method and Otsu thresholding with Frangi filter are used then to highlight tubular bacterial structures. As a result of the Frangi filter, bacterial regions are enhanced, enabling simple thresholding to accurately delineate them. With a mean Jaccard index of 0.85, the lowest mean squared error of 0.012, and the highest probabilistic Rand index of 0.78, the Otsu thresholding with Frangi filtering achieved the highest segmentation accuracy against manual annotations of over 200 microscopic images. In bio-image informatics applications, the proposed integrated pipeline demonstrates the effectiveness of tailored filtering and thresholding techniques for robust bacterial segmentation.
- 17 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
An Integrated Approach to Bacteria Structure Detection using Frangi-Thresholding Segmentation and its Impact on Analysis
Molecular characterization, classification, and other analytical tasks in microbiology require accurate segmentation of bacteria structures in microscopic images. Although bacterial shapes and textures are heterogeneous, reliable segmentation remains challenging because of low contrast, illumination variations, and variations in illumination. From microscopy scans, a pipeline for integrating image processing is presented in this work. To enhance image quality and focus on diagnostically relevant areas, an adaptive region-of-interest cropping approach is combined with Retinex enhancement, color balancing, and bi-histogram equalization. Felzenszwalb's graph-based active contour method and Otsu thresholding with Frangi filter are used then to highlight tubular bacterial structures. As a result of the Frangi filter, bacterial regions are enhanced, enabling simple thresholding to accurately delineate them. With a mean Jaccard index of 0.85, the lowest mean squared error of 0.012, and the highest probabilistic Rand index of 0.78, the Otsu thresholding with Frangi filtering achieved the highest segmentation accuracy against manual annotations of over 200 microscopic images. In bio-image informatics applications, the proposed integrated pipeline demonstrates the effectiveness of tailored filtering and thresholding techniques for robust bacterial segmentation.
- 17 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Enhancing Oral Cancer Diagnosis: IAWMF based Preprocessing in RGB and CT Images
Globally, there are over 350,000 cases of oral cancer, mainly oral squamous cell carcinomas that arise in the mouth and tongue tissues. For metastatic cases, the survival rate drops dramatically from 83% to only 65% if detected early. It is challenging, however, to identify oral lesions and precancerous conditions in their earliest stages when treatment can be most effective. CT scans and intraoral RGB photography are critical imaging methods for screening and diagnosing oral cancers. It is difficult to detect oral cancer accurately with these imaging techniques due to artifacts, noise, and poor lesion visibility. In current CAD methods, preprocessing pipelines for oral cavity images across RGB and CT are not robust or tailored. An automated, adaptive, and multimodal preprocessing pipeline is presented in this study to reduce this gap by facilitating early detection of oral cancer from RGB photographs and CT scans. A region-of-interest cropping technique is used to focus on diagnostically significant areas, specialized noise filters are used to reduce noise while maintaining tissue features, and adaptive histogram equalization is used to normalize contrast dynamically across images and highlight lesions. Oral cavity images in RGB and CT formats were used to evaluate the proposed techniques. As compared to original images, the results demonstrated significant improvements in image quality, noise reduction, lesion conspicuity, and feature visibility. Among the noise filters assessed, the Iterative Adaptive Weighted Median Filter (IAWMF) performed best with PSNR 33, SNR 28.2817, and RMSE 5.7088, which indicated the filter's effectiveness for high fidelity denoising. The median filter was also capable of reducing noise effectively with PSNR of 26.7701, SNR of 22.0518, and RMSE of 11.696. Image quality was the poorest when using Total Generalized Variation. A tailored preprocessing pipeline shows promise in aiding early diagnosis and treatment of oral cancer.
- 17 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Enhancing Oral Cancer Diagnosis: IAWMF based Preprocessing in RGB and CT Images
Globally, there are over 350,000 cases of oral cancer, mainly oral squamous cell carcinomas that arise in the mouth and tongue tissues. For metastatic cases, the survival rate drops dramatically from 83% to only 65% if detected early. It is challenging, however, to identify oral lesions and precancerous conditions in their earliest stages when treatment can be most effective. CT scans and intraoral RGB photography are critical imaging methods for screening and diagnosing oral cancers. It is difficult to detect oral cancer accurately with these imaging techniques due to artifacts, noise, and poor lesion visibility. In current CAD methods, preprocessing pipelines for oral cavity images across RGB and CT are not robust or tailored. An automated, adaptive, and multimodal preprocessing pipeline is presented in this study to reduce this gap by facilitating early detection of oral cancer from RGB photographs and CT scans. A region-of-interest cropping technique is used to focus on diagnostically significant areas, specialized noise filters are used to reduce noise while maintaining tissue features, and adaptive histogram equalization is used to normalize contrast dynamically across images and highlight lesions. Oral cavity images in RGB and CT formats were used to evaluate the proposed techniques. As compared to original images, the results demonstrated significant improvements in image quality, noise reduction, lesion conspicuity, and feature visibility. Among the noise filters assessed, the Iterative Adaptive Weighted Median Filter (IAWMF) performed best with PSNR 33, SNR 28.2817, and RMSE 5.7088, which indicated the filter's effectiveness for high fidelity denoising. The median filter was also capable of reducing noise effectively with PSNR of 26.7701, SNR of 22.0518, and RMSE of 11.696. Image quality was the poorest when using Total Generalized Variation. A tailored preprocessing pipeline shows promise in aiding early diagnosis and treatment of oral cancer.
- 15 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1, 1-6
An Elephant Identification Emissary: A Technological Odyssey in Elephant Recognition with IP & AI Solutions
With human-elephant conflicts escalating across Asia and Africa, innovative solutions are needed to identify and track individual elephants. An extensive elephant imagery dataset and advanced computer vision techniques are explored in this work for the automatic recognition of individual elephants using deep neural networks. More than 1000 images of elephants in varied natural environments were collected and processed to deal with variations in lighting, equipment, and pose. An evaluation of three convolutional neural network architectures for multi-class elephant recognition was conducted - SimCLR for self-supervised pretraining, LeNet-5, and Bidirectional Recurrent Neural Networks. With test images, the BRNN model correctly identified the elephant with 96.2% accuracy. Test results have a precision and recall of 95.5% and 96.9%, respectively. Results from these studies demonstrate the effectiveness of deep learning for elephant recognition, enabling enhanced monitoring, conservation, and conflict mitigation. However, challenges remain in model transparency, ethical implementation, integration with existing frameworks, etc. Using automated elephant identification technology, this work advances the co-existence of humans and elephants.
- 15 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1, 1-6
An Elephant Identification Emissary: A Technological Odyssey in Elephant Recognition with IP & AI Solutions
With human-elephant conflicts escalating across Asia and Africa, innovative solutions are needed to identify and track individual elephants. An extensive elephant imagery dataset and advanced computer vision techniques are explored in this work for the automatic recognition of individual elephants using deep neural networks. More than 1000 images of elephants in varied natural environments were collected and processed to deal with variations in lighting, equipment, and pose. An evaluation of three convolutional neural network architectures for multi-class elephant recognition was conducted - SimCLR for self-supervised pretraining, LeNet-5, and Bidirectional Recurrent Neural Networks. With test images, the BRNN model correctly identified the elephant with 96.2% accuracy. Test results have a precision and recall of 95.5% and 96.9%, respectively. Results from these studies demonstrate the effectiveness of deep learning for elephant recognition, enabling enhanced monitoring, conservation, and conflict mitigation. However, challenges remain in model transparency, ethical implementation, integration with existing frameworks, etc. Using automated elephant identification technology, this work advances the co-existence of humans and elephants.
- Open Access
- 24 August 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (31)
Forecasting Response Time in Elastic Cloud for Secure Resource Access in Transactional Database using Two Phase Authentication System
The key factor of IaaS in cloud computing, deals with resource sharing and efficient utilization. Though, it was accessed only by legitimate user with cloud grant and revoke mechanism that depends on cloud data management. In this paper, a two way authentication method is proposed for accessing cloud data services with transactional database to provide resource access to legitimate user. The results show the comparison between single and two way authentication method for providing scalable and consistent performance for accessing cloud data store.
- 24 August 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (31)
Open AccessForecasting Response Time in Elastic Cloud for Secure Resource Access in Transactional Database using Two Phase Authentication System
The key factor of IaaS in cloud computing, deals with resource sharing and efficient utilization. Though, it was accessed only by legitimate user with cloud grant and revoke mechanism that depends on cloud data management. In this paper, a two way authentication method is proposed for accessing cloud data services with transactional database to provide resource access to legitimate user. The results show the comparison between single and two way authentication method for providing scalable and consistent performance for accessing cloud data store.
- Open Access
- 24 March 2016
- Published by Agora University of Oradea in International Journal of Computers Communications & Control
- Vol. 11 (3) , 372-380
Forecasting Gold Prices Based on Extreme Learning Machine
In recent years, the investors pay major attention to invest in gold market ecause of huge profits in the future. Gold is the only commodity which maintains ts value even in the economic and financial crisis. Also, the gold prices are closely elated with other commodities. The future gold price prediction becomes the warning ystem for the investors due to unforeseen risk in the market. Hence, an accurate gold rice forecasting is required to foresee the business trends. This paper concentrates on orecasting the future gold prices from four commodities like historical data’s of gold rices, silver prices, Crude oil prices, Standard and Poor’s 500 stock index (S&P500) ndex and foreign exchange rate. The period used for the study is from 1st January 000 to 31st April 2014. In this paper, a learning algorithm for single hidden layered eed forward neural networks called Extreme Learning Machine (ELM) is used which as good learning ability. Also, this study compares the five models namely Feed orward networks without feedback, Feed forward back propagation networks, Radial asis function, ELMAN networks and ELM learning model. The results prove that he ELM learning performs better than the other methods.
- 24 March 2016
- Published by Agora University of Oradea in International Journal of Computers Communications & Control
- Vol. 11 (3) , 372-380
Open AccessForecasting Gold Prices Based on Extreme Learning Machine
In recent years, the investors pay major attention to invest in gold market ecause of huge profits in the future. Gold is the only commodity which maintains ts value even in the economic and financial crisis. Also, the gold prices are closely elated with other commodities. The future gold price prediction becomes the warning ystem for the investors due to unforeseen risk in the market. Hence, an accurate gold rice forecasting is required to foresee the business trends. This paper concentrates on orecasting the future gold prices from four commodities like historical data’s of gold rices, silver prices, Crude oil prices, Standard and Poor’s 500 stock index (S&P500) ndex and foreign exchange rate. The period used for the study is from 1st January 000 to 31st April 2014. In this paper, a learning algorithm for single hidden layered eed forward neural networks called Extreme Learning Machine (ELM) is used which as good learning ability. Also, this study compares the five models namely Feed orward networks without feedback, Feed forward back propagation networks, Radial asis function, ELMAN networks and ELM learning model. The results prove that he ELM learning performs better than the other methods.
- Open Access
- 4 March 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (8)
Foreign Exchange Rate Forecasting using Levenberg- Marquardt Learning Algorithm
Foreign Exchange Rate Forecasting using Levenberg- Marquardt Learning Algorithm
- 4 March 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (8)
Open AccessForeign Exchange Rate Forecasting using Levenberg- Marquardt Learning Algorithm
Foreign Exchange Rate Forecasting using Levenberg- Marquardt Learning Algorithm
- Open Access
- 14 February 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (8)
Prediction of Stock Market Price using Hybrid of Wavelet Transform and Artificial Neural Network
Prediction of Stock Market Price using Hybrid of Wavelet Transform and Artificial Neural Network
- 14 February 2016
- Published by Indian Society for Education and Environment in Indian Journal of Science and Technology
- Vol. 9 (8)
Open AccessPrediction of Stock Market Price using Hybrid of Wavelet Transform and Artificial Neural Network
Prediction of Stock Market Price using Hybrid of Wavelet Transform and Artificial Neural Network
- Open Access
Effective Features of Remote Sensing Image Classification Using Interactive Adaptive Thresholding Method
Remote sensing image classification can be performed in many different ways to extract meaningful features. One common approach is to perform edge detection. A second approach is to try and detect whole shapes, given the fact that these shapes usually tend to have distinctive properties such as object foreground or background. To get optimal results, these two approaches can be combined. This paper adopts a combinatorial optimization method to adaptively select threshold based features to improve remote sensing image. Feature selection is an important combinatorial optimization problem in the remote sensing image classification. The feature selection method has to achieve three characteristics: first the performance issues by facilitating data collection and reducing storage space and classification time, second to perform semantics analysis helping to understand the problem, and third to improve prediction accuracy by avoiding the curse of dimensionality. The goal of this thresholding an image is to classify pixels as either dark or light and evaluation of classification results. Interactive adaptive thresholding is a form of thresholding that takes into account spatial variations in illumination of remote sensing image. We present a technique for remote sensing based adaptive thresholding using the interactive satellite image of the input. However, our solution is more robust to illumination changes in the remote sensing image. Additionally, our method is simple and easy to implement but it is effective algorithm to classify the image pixels. This technique is suitable for preprocessing the remote sensing image classification, making it a valuable tool for interactive remote based applications such as augmented reality of the classification procedure.
Open AccessEffective Features of Remote Sensing Image Classification Using Interactive Adaptive Thresholding Method
Remote sensing image classification can be performed in many different ways to extract meaningful features. One common approach is to perform edge detection. A second approach is to try and detect whole shapes, given the fact that these shapes usually tend to have distinctive properties such as object foreground or background. To get optimal results, these two approaches can be combined. This paper adopts a combinatorial optimization method to adaptively select threshold based features to improve remote sensing image. Feature selection is an important combinatorial optimization problem in the remote sensing image classification. The feature selection method has to achieve three characteristics: first the performance issues by facilitating data collection and reducing storage space and classification time, second to perform semantics analysis helping to understand the problem, and third to improve prediction accuracy by avoiding the curse of dimensionality. The goal of this thresholding an image is to classify pixels as either dark or light and evaluation of classification results. Interactive adaptive thresholding is a form of thresholding that takes into account spatial variations in illumination of remote sensing image. We present a technique for remote sensing based adaptive thresholding using the interactive satellite image of the input. However, our solution is more robust to illumination changes in the remote sensing image. Additionally, our method is simple and easy to implement but it is effective algorithm to classify the image pixels. This technique is suitable for preprocessing the remote sensing image classification, making it a valuable tool for interactive remote based applications such as augmented reality of the classification procedure.