Lecture Notes in Computational Vision and Biomechanics 1...

UNCORREC

TEDPR

OOF

1 Lecture Notes in Computational Vision2 and Biomechanics

3 Volume 284

5 Series editors

6 João Manuel R. S. Tavares, Porto, Portugal7 Renato Natal Jorge, Porto, Portugal

8 Editorial Advisory Board

9 Alejandro Frangi, Sheffield, UK10 Chandrajit Bajaj, Austin, USA11 Eugenio Oñate, Barcelona, Spain12 Francisco Perales, Palma de Mallorca, Spain13 Gerhard A. Holzapfel, University of Technology, Austria14 J. Paulo Vilas-Boas, Porto, Portugal15 Jeffrey A. Weiss, Salt Lake City, USA16 John Middleton, Cardiff, UK17 Jose M. García Aznar, Zaragoza, Spain18 Perumal Nithiarasu, Swansea, UK19 Kumar K. Tamma, Minneapolis, USA20 Laurent Cohen, Paris, France21 Manuel Doblaré, Zaragoza, Spain22 Patrick J. Prendergast, Dublin, Ireland23 Rainald Löhner, Fairfax, USA24 Roger Kamm, Cambridge, USA25 Shuo Li, London, Canada26 Thomas J. R. Hughes, Austin, USA27 Yongjie Zhang, Pittsburgh, USA

Layout: T1 Standard Book ID: 442714_1_En Book ISBN: 978-3-319-71766-1

Chapter No.: FM 1 Date: 10-2-2018 Time: 7:17 pm Page: 1/12

UNCORREC

TEDPR

OOF

28 The research related to the analysis of living structures (Biomechanics) has been a source of29 recent research in several distinct areas of science, for example, Mathematics, Mechanical30 Engineering, Physics, Informatics, Medicine and Sport. However, for its successful31 achievement, numerous research topics should be considered, such as image processing32 and analysis, geometric and numerical modelling, biomechanics, experimental analysis,33 mechanobiology and enhanced visualization, and their application to real cases must be34 developed and more investigation is needed. Additionally, enhanced hardware solutions and35 less invasive devices are demanded.36 On the other hand, Image Analysis (Computational Vision) is used for the extraction of37 high level information from static images or dynamic image sequences. Examples of38 applications involving image analysis can be the study of motion of structures from image39 sequences, shape reconstruction from images, and medical diagnosis. As a multidisciplinary40 area, Computational Vision considers techniques and methods from other disciplines, such as41 Artificial Intelligence, Signal Processing, Mathematics, Physics and Informatics. Despite the42 many research projects in this area, more robust and efficient methods of Computational43 Imaging are still demanded in many application domains in Medicine, and their validation in44 real scenarios is matter of urgency.45 These two important and predominant branches of Science are increasingly considered to be46 strongly connected and related. Hence, the main goal of the LNCV&B book series consists47 of the provision of a comprehensive forum for discussion on the current state-of-the-art in these48 fields by emphasizing their connection. The book series covers (but is not limited to):

• Applications of Computational Vision andBiomechanics

• Biometrics and Biomedical Pattern Analysis• Cellular Imaging and Cellular Mechanics• Clinical Biomechanics• Computational Bioimaging and Visualization• Computational Biology in Biomedical Imaging• Development of Biomechanical Devices• Device and Technique Development for

Biomedical Imaging• Digital Geometry Algorithms for Computa-

tional Vision and Visualization• Experimental Biomechanics• Gait & Posture Mechanics• Multiscale Analysis in Biomechanics• Neuromuscular Biomechanics• Numerical Methods for Living Tissues• Numerical Simulation• Software Development on Computational

Vision and Biomechanics

• Grid and High Performance Computing forComputational Vision and Biomechanics

• Image-based Geometric Modeling and MeshGeneration

• Image Processing and Analysis• Image Processing and Visualization in

Biofluids• Image Understanding• Material Models• Mechanobiology• Medical Image Analysis• Molecular Mechanics• Multi-Modal Image Systems• Multiscale Biosensors in Biomedical Imaging• Multiscale Devices and Biomems for

Biomedical Imaging• Musculoskeletal Biomechanics• Sport Biomechanics• Virtual Reality in Biomechanics• Vision Systems

More information about this series at http://www.springer.com/series/8910



UNCORREC

TEDPR

OOF

49 D. Jude Hemanth • S. Smys50 Editors

51Computational Vision

52and Bio Inspired Computing

53

5555 123



UNCORREC

TEDPR

OOF

56 Editors58 D. Jude Hemanth59 Karunya University60 Coimbatore, Tamil Nadu61 India

62 S. Smys63 RVS Technical Campus64 Coimbatore, Tamil Nadu65 India66

67

68

69

70 ISSN 2212-939171 ISSN 2212-9413 (electronic)72 Lecture Notes in Computational Vision and Biomechanics73 ISBN 978-3-319-71766-174 ISBN 978-3-319-71767-8 (eBook)75 https://doi.org/10.1007/978-3-319-71767-876

77 Library of Congress Control Number: 201795954278

79 © Springer International Publishing AG 201880 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part81 of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,82 recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission83 or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar84 methodology now known or hereafter developed.85 The use of general descriptive names, registered names, trademarks, service marks, etc. in this86 publication does not imply, even in the absence of a specific statement, that such names are exempt from87 the relevant protective laws and regulations and therefore free for general use.88 The publisher, the authors and the editors are safe to assume that the advice and information in this89 book are believed to be true and accurate at the date of publication. Neither the publisher nor the90 authors or the editors give a warranty, express or implied, with respect to the material contained herein or91 for any errors or omissions that may have been made. The publisher remains neutral with regard to92 jurisdictional claims in published maps and institutional affiliations.

93 Printed on acid-free paper94

95 This Springer imprint is published by Springer Nature96 The registered company is Springer International Publishing AG97 The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland



UNCORREC

TEDPR

OOF

98 Contents

99 Genetic Algorithm Based Hybrid Attribute Selection Using100 Customized Fitness Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101 1102 C. Arunkumar, S. Ramakrishnan and Siva Sai Dheeraj

103 Application of Evolutionary Particle Swarm Optimization104 Algorithm in Test Suite Prioritization . . . . . . . . . . . . . . . . . . . . . . . . .105 11106 Chug Anuradha and Narula Neha

107 An Autonomous Trust Model for Cloud Integrated Framework . . . . .108 31109 C. K. Shyamala and Ashwathi Chandran

110 Combined Classifier Approach for Offline Handwritten Devanagari111 Character Recognition Using Multiple Features . . . . . . . . . . . . . . . . . .112 45113 Milind Bhalerao, Sanjiv Bonde, Abhijeet Nandedkar and Sushma Pilawan

114 Comprehensive Study on Usage of Multi Objectives in115 Recommender Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .116 55117 M. Sruthi, Sini Raj Pulari and Ramesh Gowtham

118 Visual Analysis of Genetic Algorithms While Solving 0-1119 Knapsack Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .120 68121 B. P. Sathyajit and C. Shunmuga Velayutham

122 Securing Image Posts in Social Networking Sites . . . . . . . . . . . . . . . . .123 79124 M. R. Neethu and N. Harini

125 A Types of Multi-granular Nanotopology and Its Applications . . . . . .126 92127 K. Indirani and G. Vasanthakannan

128 Early Detection of Lung Cancer Using Wavelet Feature Descriptor129 and Feed Forward Back Propagation Neural Networks Classifier . . . .130 103131 R. Arulmurugan and H. Anandakumar

v



UNCORREC

TEDPR

OOF

132 A Study on Fuzzy Weakly Ultra Separation Axioms via133 Fuzzy blB-Kernel Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .134 111135 J. Subashini and K. Indirani

136 Selection of Algorithms for Pedestrian Detection During Day and137 Night . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .138 120139 Rahul Pathak and P. Sivraj

140 Hybrid User Recommendation in Online Social Network . . . . . . . . . . .141 134142 A. Christiyana Arulselvi, S. SendhilKumar and G. S. Mahalakshmi

143 Geometrical Method for Shadow Detection of Static Images . . . . . . . .144 152145 Manoj K. Sabnis, Kavita and Manoj Kumar Shukla

146 Medical Diagnosis Through Semantic Web . . . . . . . . . . . . . . . . . . . . .147 174148 P. Monika, M. R. Vinutha, B. N. Srihari, S. Harikrishna, S. M. Sumangala149 and G. T. Raju

150 An Algorithm for Text Prediction Using Neural Networks . . . . . . . . . .151 186152 Bindu K. R., Aakash C., Bernett Orlando and Latha Parameswaran

153 A Deeper Insight on Developments and Real-Time Applications of154 Smart Dust Particle Sensor Technology . . . . . . . . . . . . . . . . . . . . . . . .155 193156 Shreya Sathyan and Sini Raj Pulari

157 Comparative Analysis of Biomedical Image Compression Using158 Oscillation Concept and Existing Methods . . . . . . . . . . . . . . . . . . . . . .159 205160 Satyawati S. Magar and Bhavani Sridharan

161 Clustering of Trajectory Data Using Hierarchical Approaches . . . . . .162 215163 B. A. Sabarish, R. Karthi and T. Gireeshkumar

164 Sentiment Analysis of Twitter Data on Demonetization Using Machine165 Learning Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .166 227167 N. M. Dhanya and U. C. Harish

168 Curvelet Based ECG Steganography for Protection of Data . . . . . . . .169 238170 Vishakha Patil and Mangal Patil

171 Kernel Based Approaches for Context Based Image Annotatıon . . . . .172 249173 L. Swati Nair, R. Manjusha and Latha Parameswaran

174 Analysis of Trabecular Structure in Radiographic Bone Images Using175 Empirical Mode Decomposition and Extreme Learning Machine . . . . .176 259177 G. Udhayakumar

178 Design of an Optimization Routing Model for Real Time Emergency179 Medical Service System in Chennai Using Fuzzy Techniques . . . . . . . .180 266181 C. Vijayalakshmi and N. Anitha

vi Contents



Curvelet Based ECG Steganographyfor Protection of Data

Vishakha Patil(&) and Mangal Patil

Deptartment of Electronics, Bharti Vidyapeeth University Collegeof Engineering, Pune, India

[email protected], [email protected]

Abstract. A ECG steganography provides safe communication over internetsuch as transformation of patient information embed on ECG signals. The datahiding technique are used here in which steganography is hiding the patientinformation in image, video, audio and signal etc. The very important challengeto ECG steganography is extract data and signal without loss. İn this technique,we use the curvelet transform for evaluating coefficients, converts informationinto binary format by using quantization and adaptive LSB algorithm forembedding. The ECG signals for steganography are taken from database ofMIT-BIH. From outcome we have observed that the signal loss is less afterembedding when the coefficients near to zero. The overlapping of embedwatermark is avoided by using sequence. The performance is observed by usingparameter such as peak signal to noise ratio, percentage residual difference,correlation and time elapsed at embedding. The extraction performance ismeasured by using correlation, BER and time elapsed parameters. At extractionbit error rate is used to calculate extract capacity. The size of patient informationis raises then BER is zero but the original signal deteroiates. Curvelet transformremoves the limitation of wavelet transform and watermarking technique. Hencethe curvelet transform is very efficient technique.

Keywords: ECG steganography � Adaptive LSB algorithm � Watermarkposition selection � Curvelet transform � Chaos encryption for embedKey encryption

1 Introduction

Everywhere health concern monitoring is enabled by recent wearable biomedicalequipment. The bio-physiological parameters are calculated by these devices and thatdata should be capable of sending via internet. The patients’ information with medicaldata is transmitted safely by using steganography. The steganography is divided intodifferent types which depends on embedding methods such as image, video, audio, etc.The personal information such as patient name, date of birth, age, address and details ofprevious treatment are store in database under personal information. This information ishidden in ECG signal which is known as ECG steganography. The ECG steganographyreduces the information loss produced at the extraction time. Most of the timesfrequency domain method is used in steganography. In this digital water-mark

© Springer International Publishing AG 2018D. J. Hemanth and S. Smys (eds.), Computational Vision and Bio Inspired Computing,Lecture Notes in Computational Vision and Biomechanics 28,https://doi.org/10.1007/978-3-319-71767-8_20

embedding, extraction and original ECG signal decomposition and also any selectedtransform method is used for signal de-composition. Generally the discrete waveletstransform (DWT) and fast discrete curvelet transform (FDCT) are used as transfor-mation methods [1–4]. Chen et al. [5] have generalized an idea of Watermarking. Fordata copyright and biological protection this technique is mostly used. They applywatermarking technique by using quantization on ECG for protection of secret data.The quantization based watermarking method is implemented by using three transformdomains such as DCT, DWT and DFT. While the watermarks embedding technique isnot invertible, the very small change occurred between the amplitude and PQRSTcomplex of ECG signal? Therefore the watermarked information can get together thenecessities of physiological diagnostics.

By using wavelet transform, the representation the curves are in limitation. Curvesare where main data such as characteristic points are there. Hence, the steganography isdeveloped by subsequent features such as (i) for analysis of the coefficients which areevaluated and which store the secret data or information (ii) remove loss duringextraction of patient information and (iii) identify and preserve original information anddecline the redundant part of the signal.

In this way, ECG steganography, by using FDCT method, is available for pro-tection of data. FDCT technique calculates the coefficients so as to re-present curves.The signal weakening is reduced by applying adaptive thresholding algorithm. Forprocessing signal taken from MIT BIH database [6], the proposed technique decreasesthe loss which occurs during extraction of information and improves the limitation ofthe watermark method. The performance of steganography algorithm is calculated byusing peak signal to noise ratio (PSNR), (MSE), percentage residual difference (PRD),correlation and time elapsed parameter at embedding side. Also at the receiver side theperformance is measured by using correlation, bit error rate (BER) and time elapsedcorrespondingly. We present the performance of the proposed method for dissimilarsizes of watermark.

2 Literature Review

Chen et al. [7] have generalized a technique is hiding Patients Confidential data in theECG Signal via Transform-Domain Quantization. The most of time the watermarkingused for data hiding and provides security to patient information. This paper theyproposes the watermarking encryption technique by using quantization and this tech-nique applied on ECG signal and provide more security. They optimize the result byimplementing DCT, DWT and DFT transform which is watermarking depends onquantization. The inverting of embed is not required in this process and this also calledthe blind watermarking. The extraction of hiding data with no understanding infor-mation of ECG signal. They verify result efficiency of this technique for result dis-cussion. They proved Transform Domain Quantization technique is very useful andefficient.

The watermark algorithm by using curvelet transform for image this idea is gen-eralized by author Xu et al. [8] In first step curvelet transform is used to decompose thecarrier image and Arnold transform is used to scramble the watermark image. In second

Curvelet Based ECG Steganography for Protection of Data 239

steps the embedding of binary image which is watermarked into coefficients of mediumfrequency. These coefficients are embed according to characteristics of human visu-alization. This method shows the performance is very good and security of system isbetter than other techniques. Noise robust is less with filtering, cropping and com-pression of jpeg is good than another attacker. The performance of scale and rotarymotion is not good for this proposed technique. They gives future scope of proposedtechnique is watermark scheme adjacent to geometric hacker using curve let transform.

Degadwala et al. [9] has proposed High Capacity Image Steganography UsingCurvelet Transform and Bit Plane Slicing. The representation of non-adaptive sparewith edges and object by using multi dimensional curvelet transform. The image isdividing into eight planes with data compressed by using bit plane slicing method. Forhiding information they used steam code RC4 method and result calculated by usingparameter mean square error and peak signal to noise ratio. They estimate in generalsecurity is high and less distortion of extracted image by using cuvelet algorithm. It isalso recognizing that robust and security is more. With the help of curvelet transformcapacity of storing information is more and without hacking information, image isextracted by using BPS. They found that the future addition of proposed work isdepends on the images size and also taken image without text. The Stream Cipher is notused in this technique they uses Block Cipher. They calculate robustness of Blurred,Rotation and other types of operation.

Jero et al. [10] they proposed ECG steganography using curvelet transform tech-nique. They uses the curvelet transform for transformation of data and quantizationused for embedding of information. The signal is divided into frequency sub bands byhelp curvelet and information is embedding in high frequency components. Thisproposed technique uses high frequency component for embedding for removing of thedeterioration and loss from signal. The extraction performance is calculated by usingbit error rate. The PSNR is increased in this technique by decreasing data size andincreasing signal size. Also they calculate the effect on a due to curvelet transform.

Ramu et al. [11] they proposed a Discrete Wavelet Transform and Singular ValueDecomposition Based ECG Steganography idea. For Secured Patient InformationTransmission The signals are decomposed by using DWT and secrete information isembedding on ECG decomposed signal by using SVD. They embed information on 2DECG image with the help of SVD. Then of information and ECG signal which issub-band format extracted by taking invert secrete data values of singular decompo-sition. They calculate the results by applying PRD, PSNR, KL and BER parameter. Thedegradation of signal of proposed method is fewer than 0.6% and this signal collectedfrom database of MIT BIH. Remember that the size of signal is more than the size ofsecretes information, so the transmission of patient data is reliable and lossless. Thedecomposition of 2D image is done by using four wavelets on transmitter and receiverside. They recognize the watermarking technique is by removing the SVD for data andsignal.

Arvind Kourav et al. [12] has proposed review on Curvelet Transform and ItsApplications. The Video surveillance, Investigation of criminal, Credit card security,Electronics banking and Forensic application used for face recognition. The face rec-ognized by using curvelet transform is very correctly without distortion. This transform

AQ1

240 V. Patil and M. Patil

is multistate transform used for Image Processing Traditional Applications. Theyestimate the how curvelet transform is better than other transform and differentapplications are used for this estimation. The wavelet transform is not good for imageinformation represented in singularities above Geometric Structures than curvelettransform. Curves on singularities for age information are designed by using curveletand singularities of point are effective in wavelet.

3 Methodology of Proposed Technique

The fundamental structural design of steganography by using ECG signal consists isdivided into two parts: Embedding (Transmitter) and Extraction (Receiver) which isshown in below Figs. 1 and 2. The ECG signal gives coefficients after any transfor-mation. The embedding side of steganography hides the secrete information in coef-ficients of signals. Remember that the size of information is less than the size of signal.After embedding we have to apply inverse transform for extraction of original ECGsignal. By using health care supplier we can receive the embedding signal. Now patientinformation is extracted by using stego signal taken at transformation and also key isused for extraction of embedding data. The watermarked method such as LSB algo-rithm coefficient matrix indicates the position of patient information in the array. Arrayis used as key which is used for security.

A. Preprocessing: (i) The one dimensional (1D) ECG signal is converted into twodimensional (2D) image.

B. Curvelet transform: With the help of curvelet transform the coefficients of image iscalculated. The curvelet transform is divided into two types such as wrappingbased fast Curvelet Transform and spaced Fast Fourier Transform. In proposedapproach we uses the wrapping based fast curvelet transform.

C. Chaos encryption: The patient data taken as secrete information that data is moresecured by using secrete key. The chaos encryption method is for key embeddingand provides more security.

D. Data Concealment: The binary stream of patient data is embed in coefficients ofECG image by using adaptive LSB algorithm.

Secret Key, Bits/sample

Input ECG signal

Curvelet transforms

DataConcealment

Secret DataStego

Chaos Encryption

Fig. 1 Data embedding (transmitter)


3.1 Preprocessing

In proposed approach the biomedical signals is taken from database of MIT BIH andchoosing signals has 128 Hz sampling rate. The Tompkins algorithm is to convert 1DECG signal into 2D ECG image. After converting signal into 2D ECG image we areapplying curvelet transform which is explained below.

3.2 Curvelet Transform

In this technique we uses the curvelet transform for transformation of 2D image. Byusing curvelet transform the coefficients of image is calculated and these are representas curved edges. The another method of fast discrete curvelet transform is unequallyspaced fast fourier transform. This method is used to calculate the curvelet coefficientsof 2D ECG image with the help of below Eq. 1.

c j; l; kð Þ ¼Z

f Shlxð Þ~Uj xð Þei b;wð Þdw ð1Þ

With the help of FFT, the FDCT converts the time domain of image into frequencydomain. The concentric square of scales that are achieved from frequency domain. Thenumbers of scales are produced from frequency space and this is measured by usinglogn. where n: size of image.

The wedges and scaled space of frequency are constructed by applying Cartesianwindow Eq. 2 is given below.

fUjðxÞ ¼ WjðxÞVj Shlxð Þ ð2Þ

where Sh ¼ 1 0� tan h 1

� �, WJ �ð Þ is radial windowpane and angular window is Vj �ð Þ.

The frequency components of image are very sensitive with coefficients. Thesefrequency components are divided into four group such as scale 1, 2, 3, 4 etc. Thescales have low frequency components, scale 2, 3 contains the components coefficientsof intermediate frequency and a high frequency coefficient is in scale 4. The binaryformat of patient information is embedding in evaluated coefficients. Hence, the

Stego Signal

DataExtraction

Text Decryption

Secret Key

Cover Signal and Data

Fig. 2 Data extraction (receiver)


coefficient of scale 4 is modified for preservations the patient data. The scale 4 group isvery good for embedding information because these remove the deterioration of dataand 2D image.

3.3 Select Adaptive LSB Position

In this LSB position, the overlapping of LSB distribution and coefficient valuesmodification is avoided. The number of adaptive LSB position represented in Fig. 6.The LSB location and its neighbors are embedded in watermark bit. However, thecapacity of watermarking is reduced in this technique. Generally, the work out ofsequence n*n is in sightless way.

For embedding the binary format of patient information is needed and that format ishides in CD by using quantization Eq. 3 shown below

C� i; jð Þ ¼ a CD i; jð Þ�� w: ð3Þ

3.4 Watermark Extractions

For extraction of patient information ECG image is put in FDCT. The watermarkedcoefficients are finding by using key. Apply inverse watermark embedding and patientdata are extracted with the help of Eq. 4 where

Wr ¼ 1; if c i; jð Þ[ 00 if c i; jð Þ\0

ð4Þ

Here we apply the adaptive LSB algorithm and achieves less signal deterioration. Thisoutcome inside no concession going on diagnosibility with good informationextraction.

4 Results and Discussion

We taken four different ECG signals and four different secrete biometric images forobservation of parameters. This embedding done with text data (Patient Confidentialinformation: Name: X.Y.Z, Date of Birth: 15/6/1994 Address: Pune Medicare Number:9890124814 Telephone Number: 1234567 Patient Diagnoses information 1. BloodPressure. 2. Temperature. 3. Glucose Level) is same for all signals. The PSNR, MSE,correlation, PRD and time elapsed is calculated for each signal. The data hidingcapacity is depends on ECG signal size and number of bits stored per ECG sample.Below Table 1 shows the input signal secrete image and embed signal such as stegosignal. Table 2 shows parametric evaluation of these signals after embedding.

AQ2


4.1 After Varying the Bits Per Sample Effect on Stego Signaland Parameters Is Given Below

The ECG signal 1 is taken for hiding secrete data such as secrete image1 and text(Patient Confidential information: Name: X.Y.Z, Date of Birth: 15/6/1994 Address:Pune Medicare Number: 9890124814 Telephone Number: 1234567 Patient Diagnosesinformation (1) Blood Pressure. (2) Temperature. (3) Glucose levels). The 1 and 2 bits

Table 1 Difference between input and embed signal

Bits Input signals Secrete image Stego signal

3 ECG signal 1 Sec img1




Table 2 Parameter evaluation of above table at embedding side

Parameters ECG signal1,img1

ECG signal2,img2

ECG signal3,img3

ECG signal4,img4

PSNR(db) 60.28 59.9862 61.2277 61.3718MSE 8.2925 9.3565 9.10725 8.6285Correlation 0.999887 0.999825 0.999962 0.999862PRD 0.00113096 0.00119855 0.00118666 0.00115305Timeelapsed

1.16463 0.490931 0.485996 0.484288


hiding per ECG sample is not possible for this steganography and the 3bits per sample isvery good than others. The number of bits per sample is increases then the PSNR isreduced, MSE is increased, correlation is decreases and PRD is increases which areshown in table and graph. The effect of increased bits per sample on stego signal is givenbelow table a. After extraction process the secrete data is extracted completely and thisproven by parameter correlation and BER. The correction value is ‘1’ for all bits’ persample hence the correlation value 1 is indicates the all data are extracted completely.The BER value at extraction side is zero. The number of increasing bits per sampleverses PSNR and correlation graphs has shown in Figs. 3 and 4; (Tables 3 and 4).

Fig. 4 No of bits/sample versus correlation

Fig. 3 No of bits/sample versus PSNR


Table 3 Effect of rising bits/sample on signal

For 3 bits/sample For 3 bits/sample





Table 4 After varying bits effect on parameter

No. of bits/sample PSNR (db) MSE Correlation PRD (%) Time elapsed

3 60.2662 8.31875 0.999886 0.00113772 0.4611944 55.8591 22.9492 0.999698 0.00163382 0.3046115 50.3435 81.7205 0.998885 0.00279256 0.4849666 45.2571 263.614 0.996595 0.00451624 0.4830197 40.2338 838.104 0.988711 0.00750812 0.4689578 35.1577 2697.16 0.967 0.0120463 0.4891439 28.1867 13428 0.87288 0.0297179 0.45770110 22.4296 50550.1 0.74025 0.567817 0.46400711 15.0723 275077 0.53394 0.3226 0.47924612 16.3363 205612 0.294792 0.143462 0.493095


4.2 Parameter Evaluation After Extraction Data

The correlation for all variable bits is one and BER is zero but signal deterioration isrises. The data extraction is completely done by using curvelet transform than wavelettransform is proven from this parameter. The time elapsed is varying by increasingnumber of bit’s per sample this shown in below graph c (Fig. 5).

5 Conclusions

Execution of curvelet based ECG steganography calculation utilizing versatile LSBsystem was considered in this work. At the point when coefficients are adjusted, thecover flag falls apart and influence diagnosability. It is watched that for around 1.3times increment in mystery information measure is expanded, the flag decay is around10%. We contemplate impact of number of bits per test on all parameters. The 0% BERfor all watermark sizes demonstrates that proposed approach recovers the mysteryinformation with no misfortune. PRD of proposed approach is right around zero for allwatermark sizes. It is watched that the PSNR esteem is great all through whichguarantees the indistinctness of watermark. For increment in watermark estimate, thereis diminish in the PSNR esteem. Along these lines proposed philosophy can be utilizedfor dependable steganography.

Acknowledgements. I am greatly pleased to proof: M. V. Patil who guided me through myproject curvelet based ECG steganography for protection of data.

References

1. Ibaida, A., Khalil, I.: Wavelet-based ECG steganography for protecting patient confidentialinformation in point-of-care systems. In: IEEE Trans. Biomed. Eng, pp. 3322–3330. (2013)

Fig. 5 No of bits/sample versus time elapsed extraction

AQ3

AQ4


2. Miyara, K., Chen, F., Nakao, Z.: Digital watermarking based on curvelet transform. NinthInt. Symp. Signal Process. (2007)

3. Taby, A., Naima Fiete, F.: High capacity image steganography based on curvelet transform.In: Proc. Fourth Int. Conf. Dev. Systems Eng., pp. 191–196 (2011)

4. Ji, F., Huang, D., Deng, C., Zhang, Y., Miao, W.: Robust curvelet-domain imagewater-marking based on feature matching. Int. J. Comp. Math. 88, 3931–3941 (2011)

5. Chen, S.T., Guo, Y.J., Huang, H.N., Kung, W.M., Tseng, K.K., Tu, S.Y.: Hiding patientsconfidential data in the ECG signal via transform-domain quantization scheme. J. Med. Syst.(2014)

6. Moody, G.B., Mark, R.: MIT-BIH arrhythmia database directory. In: MIT-BIH DatabaseDistribution, Harvard-MIT Division of Health Sciences and Technology, MassachusettsInstitute of Technology. Available from http://www.physionet.org/physiobank/database/html/mitdbdir/mitdbdir.htm (1992)

7. Chen, S.-T., Guo, Y.-J., Huang, H.-N., Kung, W.-M., Tseng, K.-K., Tu, S.-Y.: Hidingpatients confidential data in the ECG signal via transform-domain quantization scheme.J. Med. Syst. 38–54. doi:https://doi.org/10.1007/s10916-014-0054-9 (2014)

8. Xu, J., Pang, H., Zhao, J.: Digital image watermarking algorithm based on fast curvelettransform. Software Engineering & Applications, 939–943. doi:https://doi.org/10.4236/jsea.2010.310111 (2010)

9. Degadwala, S., Thakkar, A., Nayak, R.: High capacity ımage steganography using curvelettransform and bit plane slicing. Int. J. Adv. Res. Comp. Sci. ISSN No. 0976–5697, 4 (2013)

10. Jero, E., Ramu, P., Ramakrishnan, S.: ECG steganography using curvelet transforms.Biomed. Sig. Process. Control 22, 161–169. doi:https://doi.org/10.1049/el.2015.3218

11. Jero, S.E., Ramu, P., Ramakrishnan, S.: Discrete wavelet transform and singular valuedecomposition based ECG steganography for secured patient ınformation transmission.J. Med. Syst. 38–132. doi:https://doi.org/10.1007/s10916-014-0132-z (2014)

12. Kourav, A., Singh, P.: Review on curvelet transform and its applications. Asian. J. Electr.Sci. ISSN 2249–6297. 2(1) pp. 9–13, (2013)

AQ5


http://www.physionet.org/physiobank/database/html/mitdbdir/mitdbdir.htm

http://www.physionet.org/physiobank/database/html/mitdbdir/mitdbdir.htm

http://dx.doi.org/10.1007/s10916-014-0054-9

http://dx.doi.org/10.4236/jsea.2010.310111

http://dx.doi.org/10.4236/jsea.2010.310111

http://dx.doi.org/10.1049/el.2015.3218

http://dx.doi.org/10.1007/s10916-014-0132-z

Author Query Form

Book ID : 442714_1_En

Chapter No : 20123

the language of science

Please ensure you fill out your response to the queries raised belowand return this form along with your corrections.

Dear Author,During the process of typesetting your chapter, the following queries havearisen. Please check your typeset proof carefully against the queries listed belowand mark the necessary changes either directly on the proof/online grid or in the‘Author’s response’ area provided below

Query Refs. Details Required Author’s Response

AQ1 The author names cited with the references seems to be mismatch with thereference list. Please check and confirm.

AQ2 Please note that there is no Fig. 6 in this Chapter, but Fig. 6 is cross linked in thesentence 'The number of adaptive LSB position represented in Fig. 6'. Kindlyconfirm.

AQ3 Please confirm if the section headings identified are correct.

AQ4 Please check and confirm if the inserted citations of Fig. 5 and Tables 3, 4 arecorrect. And kindly note that Figures are renumbered to maintain the sequentialorder. If not, please suggest an alternate citations. Please note that figures andtables should be cited sequentially in the text.

AQ5 Please supply the year of publication for Reference [11].

Kernel Based Approaches for Context BasedImage Annotatıon

L. Swati Nair(&), R. Manjusha, and Latha Parameswaran

Department of Computer Science Engineering, Amrita School of Engineering,Amrita Vishwa Vidyapeetham, Coimbatore, India

[email protected], [email protected],

[email protected]

Abstract. The Exploration of contextual information is very important for anyautomatic image annotation system. In this work a method based on kernels andkeyword propagation technique is proposed. Automatic annotation with a set ofkeywords for each image is carried out by learning the image semantics. Thesimilarity between the images is calculated by Hellinger’s kernel and RadialBias Function kernel(RBF)kernel. The images are labelled with multiple key-words using contextual keyword propagation. The results of using the twokernels on the set of features extracted are analysed. The annotation resultsobtained were validated based on confusion matrix and were found to have agood accuracy. The main advantage of this method is that it can propagatemultiple keywords and no definite structure for the annotation keywords has tobe considered

Keywords: Automatic image annotation � Hellinger’s kernel � RBF kernelSemantics � Contextual keyword propagation � Gabor features � Haralickfeatures

1 Introduction

Due to the rapid growth of archiving of images, the need for indexing and searchingimages effectively has increased significantly today. In spite of the fact that manyContent Based Image Retrieval methods are prevalent, searching based on imagefeature is rather difficult for the users. Most of the users prefer searching with textualqueries. This can be achieved by annotating the images manually and then searchingthe annotated images using textual queries. But it is a known fact that manual anno-tation of a large number of images is very much time consuming, expensive and alsoinvolves considerable efforts. Hence, automatic image annotation methods are pre-ferred over manual annotation for efficient retrieval of images. Thus, automatic imageannotation with keywords is widely used which involves the learning of semantics ofimages. Automatic image annotation cannot be accurate if the context of the scene isnot taken into account for any object in the scene. So context based image annotationbecomes a very important aspect.

© Springer International Publishing AG 2018D. J. Hemanth and S. Smys (eds.), Computational Vision and Bio Inspired Computing,Lecture Notes in Computational Vision and Biomechanics 28,https://doi.org/10.1007/978-3-319-71767-8_21

Lecture Notes in Computational Vision and Biomechanics 1...

Documents

Transcript of Lecture Notes in Computational Vision and Biomechanics 1...