Time filter

Source Type

Mali S.N.,Sinhgad Institute of Technology and Science
11th IEEE India Conference: Emerging Trends and Innovation in Technology, INDICON 2014 | Year: 2015

Robust, secured, high embedding capacity and invisible digital image watermarking techniques are key requirements in the world of information security. Many existing image watermarking techniques fail to achieve perceptual quality and robustness simultaneously under high payload scenario promising strong security, because these requirements conflict each other. This paper presents non-blind digital image watermarking technique and its comparative performance analysis in discrete wavelet transform (DWT) and DWT-Fast Walse Hadamard transform-Singular value decomposition (DWT-FWHT-SVD) domains. The simple, symmetric and orthogonal 'Haar' wavelet is used for DWT decomposition and Fibonacci-Lucas transform along with affine transforms are used to implement embedding security. The experimental results show better performance in DWT-FWHT-SVD domain than DWT domain for candidate images of 512×512 size namely Lena, peppers, baboon, lake and plane and grey scale watermark of 256×256 sizes. In DWT-FWHT-SVD domain, better perceptual quality is achieved with Lena image giving peak signal to noise ratio(PSNR)=75.8446, peppers image with PSNR=82.8343, baboon image with PSNR=94.0575, lake image with PSNR= 86.9526 and plane with PSNR=89.9170 with significant quality of extracted watermark. For all test images with varying scale factor in the range of K=0.25 to 0.55, PSNR and normalized corelation(NC) values are better with compared to DWT domain. As DWT-FWHT-SVD based results found superior, we further focused on DWT-FWHT-SVD domain and demonstrated that our technique is robust against 19 various noise addition and filtering attacks. The DWT-FWHT-SVD domain technique is also found superior than existing redundant-DWT-SVD based method against 12 different attacks. © 2014 IEEE. Source

Bhosale K.,Vishwaniketan Institute of Management Entrepreneurship Engineering and Technology | Rohokale V.,Sinhgad Institute of Technology and Science
2015 International Conference on Pervasive Computing: Advance Communication Technology and Application for Society, ICPC 2015 | Year: 2015

With the help of digital resources, self-description of smart objects causes some limitation into the input qualities because of the long established physical appearances. Also to realize the capability of objects to understand and recognize the data. If an uninstrumented surface is provided, the object should be able to sense gestures and touches upon the surface. To design an area for supporting whole body metaphors physical movements and position of entire body.Objects such as sensors, processors and radios are integrated invisibly into smart sensitive objects for user interaction. In this paper, we propose to rectify the excitation and response imbalance by augmentation of smart objects with natural appearance of it. To achieve this, we propose the implementation of computer vision for cooperative augmented reality by open CV using matlab. © 2015 IEEE. Source

Mali S.N.,Sinhgad Institute of Technology and Science
IET Conference Publications | Year: 2012

'Secure Voting System' is heart of any democracy. There are number of nationwide voting system adopted all over the world, but each of them has their own shortfalls. The remote internet voting systems still suffer many problems. These are reasons, why manual voting is still in practice in many developing and developed nations in this internet era also. Thus, complete, strongly secured and user friendly 'E-Voting System' is need of time. The aim of this paper is to present multilayer secured, internet based voting system using biometric and wavelet based image watermarking. Strongly secured watermarking technique for voter's color photograph in YCgCb color space is processed by embedding voter's fingerprint as watermark. The watermark embedding is done securely through number of levels. This technique yields Peak Signal to Noise Ratio (PSNR) up to 54.26 and Normalised Correlation (NC) equals to 1 indicating exact recovery of fingerprint. The complete system is maintained 'user friendly'. Source

Salunkhe U.R.,NBN Sinhgad Technical Institute Campus | Mali S.N.,Sinhgad Institute of Technology and Science
2015 International Conference on Advances in Computing, Communications and Informatics, ICACCI 2015 | Year: 2015

During the last few years, imbalanced data classification issue has gained a great deal of attention. Many real life applications suffer from imbalanced distribution of data that can be handled by using different approaches such as data level, algorithm level or classifier ensembles. Single level as well as multi level classifier ensemble technique has shown improvement in classification performance. Also data level approaches are independent of classifier being used. In past few years, combination of data level and classifier ensemble technique has been applied and has proved to be effective. This paper explores the impact of pre-processing algorithm on the performance of classifier ensemble approach for imbalanced data set. The aim of this study is to investigate the effect of pre-processing on two level classifier ensemble approaches. Experimental work and analysis of results shows that use of pre-processing is not beneficial for Random Subspace Method since results reflect performance degradation while AdaBoost has shown improvement due to application of pre-processing. © 2015 IEEE. Source

Salunkhe U.R.,University of Pune | Mali S.N.,Sinhgad Institute of Technology and Science
Procedia Computer Science | Year: 2016

Imbalanced learning for classification problems is the active area of research in machine learning. Many classification systems like image retrieval and credit scoring systems have imbalanced distribution of training data sets which causes performance degradation of the classifier. Re-sampling of imbalanced data is commonly used to handle imbalanced distribution as it is independent of the classifier being used. But sometimes they can remove necessary data of the class or can cause over-fitting. Classifier Ensembles have recently achieved more attention as effective technique to handle skewed data. The focus of the work is to gain advantages of both data level and classifier ensemble approach in order to improve the classification performance. We present a novel approach that initially applies pre-processing to the imbalanced dataset in order to reduce the imbalance between the classes. The pre-processed data is provided as training dataset to the classifier ensemble that introduces diversity by using different training datasets as well as different classifier models. The experimentation conducted on the eight imbalanced datasets from KEEL repository helps to prove the significance of the proposed method. A comparative analysis shows the performance improvement in terms of Area under ROC Curve (AUC). © 2016 The Authors. Published by Elsevier B.V. Source

Discover hidden collaborations