DEBASISH PRADHAN
@diat.ac.in
Associate Professor, School of Computer Engneering & Mathematical Sciences
RESEARCH, TEACHING, or OTHER INTERESTS
Computational Mathematics
Scopus Publications
Scopus Publications
Swapnil Kale, Debasish Pradhan, and Madhusmita Tripathy
Springer Science and Business Media LLC
Pratibha Bhandari, Neelesh S Upadhye, and Debasish Pradhan
IEEE
The paper acknowledges the complex behaviour of time series data, renowned for its inherent chaotic nature and inclination towards sudden, significant high jumps. To address such intricate data, the incorporation of stochastic differential equations (SDEs) with noise can be done to accurately represent the data and its inherent traits. To better comprehend the distribution sustained by the data, the paper explores the realms of stable distribution, particularly relevant for data exhibiting heavier tails. Therefore, this review highlights the utility of modelling white noise within the frame work of a non-Gaussian distribution governed by Lévy motion. In this context, it proposes using alpha stable Lévy motion due to its remarkable ability to closely mirror the data's characteristics. Thus harmonizing the noise modelling process with the data's intricate dynamics. It aims at the techniques proposed in various research papers and discerning the most effective approach for analyzing financial time series data. Additionally, it also outlines the major improvements and future implementations in this domain.
B R Kapuriya, Debasish Pradhan, and Reena Sharma
Springer Science and Business Media LLC
Swapnil Kale and Debasish Pradhan
Elsevier BV
Ajeet Kumar Verma, Anindya Pain, Ekansh Agarwal, and Debasish Pradhan
Springer Science and Business Media LLC
Swapnil Kale and Debasish Pradhan
Springer Science and Business Media LLC
Nitish Patel and Debasish Pradhan
IEEE
The coronavirus 2019 is a worldwide pandemic declared by the world health organization (WHO). It starts in China, Wuhan in November 2019 and spread all over the world. As time passed, the detection and clinical treatment of COVID-19 is developed by the researchers. COVID-19 is detected using a reverse transcription-polymerase chain reaction (RT-PCR) test, which is precise but requires two days to complete. Hence, the researchers proposed many classification models, which are mainly based on artificial intelligence. Mainly these classification models are using chest X-ray images for the detection of COVID-19. In this paper, we proposed a deep convolutional neural network model architecture to classify chest X-ray images. We called this model the base model, which is the first train to classify normal and abnormal chest X-ray images. Using the transfer learning technique, we retrained this model for four-classes classification (i.e., Normal, COVID-19, Pneumonia, and Pneumothorax). We obtain 73.9% accuracy for the base model (i.e., binary classification) and 83.2% accuracy for fine-tuned model (i.e., four-classes classification).
Viwek Mahto, Debasish Pradhan, and Rajendra Shrikant Deodhar
Springer Singapore
Deepa Chakravarty and Debasish Pradhan
Springer Science and Business Media LLC
Deepa Chakravarty and Debasish Pradhan
IEEE
Image segmentation is an important topic in the field of computer vision and image processing. It is also of major significance in the area of medical imaging. Segmenting medical images with thin tube-like structures is a major challenge. The model proposed in this paper can be used to obtain a desired segmented result from an image having vessel or tubular features. This model uses the concept of solving Mumford-Shah model to obtain a smooth image. This smoothened image is segmented using the idea of Tight-Frame-based Algorithm with Eigenvector (TFAE). This hybrid method gives better segmented results than that of TFAE method alone. The given model also works well with noised images. The proposed method has been implemented on images with and without noise. The result obtained by this method has been further compared with existing models.
B R Kapuriya, Debasish Pradhan, and Reena Sharma
IEEE
Selective image segmentation extracts object of interest from an image based on user input. There are many variational methods which are effective in segmenting object which has uniform intensity. But many times object with multiple intensities needs to extracted. In this paper, we propose an active contour based algorithm for selective segmentation of objects in vector valued image. Here, we introduced new energy function by adding new geometric constraints based fidelity terms from local region of the object for vector valued images. The model minimizes new functional over the length of the contour along various parts of the object in different component images. Experimental results shows that the proposed method is effective in segmenting object having multi intensities. Performance of the proposed method is determined using Jaccard’s similarity index for all methods.
Swapna Jawalkar, Odelu Ojjela, and Debasish Pradhan
Springer Singapore
B. R. Kapuriya, Debasish Pradhan, and Reena Sharma
Springer International Publishing
B. R. Kapuriya, Debasish Pradhan, and Reena Sharma
Springer Science and Business Media LLC
Rutuja V. Savant and Debasish Pradhan
IEEE
This article deals with the analysis of noise parameter estimation for captured image under motion blurring and illumination flickering. A captured image is first analysed for determination of motion blur parameters i.e., direction of blur and pixel displacement length (PDL) and it is carried out using edge information of the image. Subsequently a blur-free image is processed for estimation of illumination flicker noise parameters and then for the estimation of camera parameters-camera gain, mean and variance of Gaussian white noise and illumination flickering co-efficient.
Sunita V. Dhavale, Rajendra S. Deodhar, Debasish Pradhan, and L.M. Patnaik
Inderscience Publishers
Amrit Panda and Debasish Pradhan
IEEE
In this paper we concentrate on understanding the Hyperspectral Image subspace, spectral processing of the Hyperspectral Image using Spectral Angle Mapping to achieve target detection. A combined spatial-spectral integrated processing algorithm is proposed to be implemented in cases where spectral processing produces probable target pixels that are spatially spread. Atmospheric error correction is done using the method of Internal Average Relative Reflectance. To reduce processing time necessary dimensionality reduction has been implemented using Principal Component Analysis. EO-1 Hyperion datasets have been used for this project. The results of both the spectral classification and the proposed integrated spatial-spectral processing algorithm with and without atmospheric error correction as well as with and without dimensionality reduction has been analysed using ENVI Image processing toolbox as well as using MATLAB. The effectiveness of each method and the difference in results using different platforms has been inferred from the numerical experiments.
Nilam A. Chirame, Debasish Pradhan, and S.D. Naik
Elsevier BV
Sunita V. Dhavale, R. S. Deodhar, Debasish Pradhan, and L. M. Patnaik
Informa UK Limited
ABSTRACT In this paper, we propose a new state transition based embedding (STBE) technique for audio watermarking with high fidelity. Furthermore, we propose a new correlation based encoding (CBE) scheme for binary logo image in order to enhance the payload capacity. The result of CBE is also compared with standard run-length encoding (RLE) compression and Huffman schemes. Most of the watermarking algorithms are based on modulating selected transform domain feature of an audio segment in order to embed given watermark bit. In the proposed STBE method instead of modulating feature of each and every segment to embed data, our aim is to retain the default value of this feature for most of the segments. Thus, a high quality of watermarked audio is maintained. Here, the difference between the mean values (Mdiff) of insignificant complex cepstrum transform (CCT) coefficients of down-sampled subsets is selected as a robust feature for embedding. Mdiff values of the frames are changed only when certain conditions are met. Hence, almost 50% of the times, segments are not changed and still STBE can convey watermark information at receiver side. STBE also exhibits a partial restoration feature by which the watermarked audio can be restored partially after extraction of the watermark at detector side. The psychoacoustic model analysis showed that the noise-masking ratio (NMR) of our system is less than −10dB. As amplitude scaling in time domain does not affect selected insignificant CCT coefficients, strong invariance towards amplitude scaling attacks is also proved theoretically. Experimental results reveal that the proposed watermarking scheme maintains high audio quality and are simultaneously robust to general attacks like MP3 compression, amplitude scaling, additive noise, re-quantization, etc.
Manorama Pokheriya and Debasish Pradhan
IEEE
Object detection and tracking plays an important role in the field of video surveillance and has been discussed since many years. There are several techniques available in literature. But to find out a robust method which can give the better result is a challenging job. In this paper, we proposed a method which can detect and track the motion of an object. The proposed method is a combination of adaptive background subtraction, a trained silhouette based model for detection and Kalman filter for tracking purpose.
Milan Patnaik, Ajay Mathew, M. S. Gill, and Debasish Pradhan
IEEE
This paper proposes a fast and robust text-independent speaker identification system for all types of radio networks. The radio-conversations contain speech from various speakers along with radio noise. A novel approach to segment the radio-conversations into speaker homogenous speech segments named as Reciever Noise Segmentation (RxNSeg) is proposed which first identifies the receiver radio-noise and then finds the boundaries for speaker homogeneous speech segments in the radio-conversation. Various techniques for clustering of speech segments to arrive at speaker homogenous clusters to train speaker models are evaluated. A novel top-down approach named as Find One Long Speech Segment (FOLSS) for finding at least one long speaker homogenous segment for each speaker present in a radio-conversation is proposed in lieu of traditional clustering techniques. Speaker modeling using Gaussian Mixture Model (GMM) and adapted-GMM are considered. The two speaker modeling methods with proposed RxNSeg and FOLSS show an average 86:32% reduction in testing time without significant loss of speaker identification accuracy as com-pared to traditional segmentation and clustering techniques.
S.V. Dhavale, L.M. Patnaik, R.S. Deodhar, and D. Pradhan
Institution of Engineering and Technology
This paper presents a robust, stereo audio signal watermarking algorithm. The inter-channel redundancy of stereo audio signal is exploited by calculating average and difference signals from the LR channels of stereo audio. Discrete Wavelet Transform (DWT) followed by Discrete Cosine Transform (DCT) are applied to both average and difference signals. Few selected transform domain coefficients of average signal are used for embedding a robust watermark along with synchronization codes. A semi-fragile watermark is embedded for integrity check in a few of the transform domain coefficients of difference signal. Quantization technique is used for embedding. As both average and difference signals are used for embedding, the watermark will spread evenly in both LR channels without interfering each other. As more signal energy is concentrated in average signal compared to difference signal, larger quantization step size can be used for embedding watermark in average signal. As both the channels are used for embedding, high data payload is achieved. The proposed scheme is blind where both the watermarks are extracted independently without using the original audio. Extensive experimental results assure that the proposed watermarking scheme provides high SNR and is highly robust to common attacks like, MP3 compression, cropping, time shifting, filtering, resampling and re-quantization.
Debasish Pradhan, Baskar Shalini, Neela Nataraj, and Amiya K. Pani
Springer Science and Business Media LLC