manami barthakur | Tezpur University (original) (raw)
Uploads
Papers by manami barthakur
Data Visualization and Knowledge Engineering
Satellite images carry essential information required for range of applications. Information extr... more Satellite images carry essential information required for range of applications. Information extraction from satellite images is a challenging issue and requires a host of support. Further automation of information extraction, reliability and decision making regarding content are essential and vital elements. Of late, among the learning based approaches deep neural network (DNN) supported methods have been to efficient and reliable. In this work a DNN based approach optimized for satellite images have been discussed. The chapter includes the details of the techniques adopted, including the DNN topologies, experiments performed, results and related discussion. The work includes details of a specially configured and trained CNN topology which is found to be suitable for satellite image segmentation. Finally an approach based on deep learning for semantic segmentation in satellite images is also proposed. From the experimental results it was seen that the method is appropriate for real world and reliable.
IEEE Access, 2020
Content extraction from satellite images continues to evolve with the application of learning aid... more Content extraction from satellite images continues to evolve with the application of learning aided approaches. Recently, with the addition of deep learning (DL) based methods, content extraction from satellite images has become more reliable and efficient, yet challenges continue to exist as these methods require a large number of training and annotated images to enable effective learning by these networks. For high-resolution satellite images, limited training data is a familiar problem. Therefore, amongst the DL-based methods, semi-supervised adversarial approaches represent an emerging area of application in content extraction from satellite images. Semi-supervised adversarial methods adopt a combination of unsupervised training and labeled data to process applied inputs to generate reliable classification. In this paper, a semi-supervised adversarial learning method, which includes architectural expansion and several other additions, is reported that is used for content extract...
2019 2nd International Conference on Innovations in Electronics, Signal Processing and Communication (IESC)
International Journal of Computer Applications, 2013
Multiplicative speckle noise which is inherently present in medical ultrasound images degrades th... more Multiplicative speckle noise which is inherently present in medical ultrasound images degrades the important clinical informations and badly affects the quality of the diagnosis. It is necessary to reduce the speckle noise to improve the visual quality of ultrasound images for better diagnoses. In this paper, a wavelet based method for despeckling of the ultrasound images is introduced where a local Wiener filter along with speckle reducing anisotropic diffusion (SRAD) filter are employed in a homomorphic framework. The signal variance in the local wiener filter is estimated from the output image of the SRAD filter. Since the size and shape of the locally adaptive window is an important issue in estimating the signal variance, nearly arbitrarily shaped windows are used for better performance. The experimental results using synthetically speckled ultrasound images show that the speckle noise is reduced to a great extent while preserving the important clinical information. In order to demonstrate the effectiveness of the proposed method, the method is compared with several other existing methods in terms of peak signal to noise ratio (PSNR), structural similarity index (SSIM), edge preservation index (β), and standard deviation to mean (S/M) ratio.
In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of vari... more In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.
Studies in Computational Intelligence, 2012
In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of vari... more In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.
Data Visualization and Knowledge Engineering
Satellite images carry essential information required for range of applications. Information extr... more Satellite images carry essential information required for range of applications. Information extraction from satellite images is a challenging issue and requires a host of support. Further automation of information extraction, reliability and decision making regarding content are essential and vital elements. Of late, among the learning based approaches deep neural network (DNN) supported methods have been to efficient and reliable. In this work a DNN based approach optimized for satellite images have been discussed. The chapter includes the details of the techniques adopted, including the DNN topologies, experiments performed, results and related discussion. The work includes details of a specially configured and trained CNN topology which is found to be suitable for satellite image segmentation. Finally an approach based on deep learning for semantic segmentation in satellite images is also proposed. From the experimental results it was seen that the method is appropriate for real world and reliable.
IEEE Access, 2020
Content extraction from satellite images continues to evolve with the application of learning aid... more Content extraction from satellite images continues to evolve with the application of learning aided approaches. Recently, with the addition of deep learning (DL) based methods, content extraction from satellite images has become more reliable and efficient, yet challenges continue to exist as these methods require a large number of training and annotated images to enable effective learning by these networks. For high-resolution satellite images, limited training data is a familiar problem. Therefore, amongst the DL-based methods, semi-supervised adversarial approaches represent an emerging area of application in content extraction from satellite images. Semi-supervised adversarial methods adopt a combination of unsupervised training and labeled data to process applied inputs to generate reliable classification. In this paper, a semi-supervised adversarial learning method, which includes architectural expansion and several other additions, is reported that is used for content extract...
2019 2nd International Conference on Innovations in Electronics, Signal Processing and Communication (IESC)
International Journal of Computer Applications, 2013
Multiplicative speckle noise which is inherently present in medical ultrasound images degrades th... more Multiplicative speckle noise which is inherently present in medical ultrasound images degrades the important clinical informations and badly affects the quality of the diagnosis. It is necessary to reduce the speckle noise to improve the visual quality of ultrasound images for better diagnoses. In this paper, a wavelet based method for despeckling of the ultrasound images is introduced where a local Wiener filter along with speckle reducing anisotropic diffusion (SRAD) filter are employed in a homomorphic framework. The signal variance in the local wiener filter is estimated from the output image of the SRAD filter. Since the size and shape of the locally adaptive window is an important issue in estimating the signal variance, nearly arbitrarily shaped windows are used for better performance. The experimental results using synthetically speckled ultrasound images show that the speckle noise is reduced to a great extent while preserving the important clinical information. In order to demonstrate the effectiveness of the proposed method, the method is compared with several other existing methods in terms of peak signal to noise ratio (PSNR), structural similarity index (SSIM), edge preservation index (β), and standard deviation to mean (S/M) ratio.
In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of vari... more In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.
Studies in Computational Intelligence, 2012
In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of vari... more In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.