International Journal of Recent Innovation in Engineering and Research Scientific Journal Impact Factor - 3.605 by SJIF e- ISSN: 2456 2084 SUITABLE IMAGE RETRIEVAL FOR IOT APPLICATION V.Krishnasree 1 and T.Divija 2 1 Prof, ECE, VNRVJIET, BHACHUPALLY-HYD, INDIA 2 M.Tech, ES, VNRVJIET, BHACHUPALLY-HYD, INDIA Abstract-The proposed system is mainly considered for big data images which are obtained from different sensor/devices which have to be uploaded to the main server. The sensors/devices are mainly used in medical streams, saving the big data in database requires more storage capacity to overcome this problem we use fusion and compression techniques. The Retrieval of image is reverified with the samples of the big data and thus we can avoid problems such as memory storage etc. Image Fusion is a technique of combining the useful information from a set of images into a single image, where the output fused image will be more informative and useful than any of the input images. Image fusion techniques can improve the quality and increase the application area of these data. This research paper compares the experimental results generated by the proposed Reduced SVD based image retrieval model with the standard method of image retrieval. The proposed Reduced SVD based image retrieval model is then transmitted to the main server using IOT Devices/techniques. We also compare different compression techniques and choose the best fit technique for image retrieval process. Keywords Image Retrieval, Image fusion, Images, IOT application, Reduced SVD. I. INTRODUCTION Huge amount of data in sense called Big-Data which is obtained from different sensors/devices have to be uploaded to the main server, the transmission of this data to the main server is referred to as Internet of Things (IOT). While uploading and storing the bulk images we could face many issues [1][10] such as transmission errors, high utility of bandwidth, storage insufficiency, bulk amount of images requires much time for uploading to the destination and hence time increases. To overcome the issues mentioned above we can use Compression/Fusion techniques. Fusion techniques are mainly applied for similar images such as video frames/ Camera images, these images are fused using different fusion techniques [1] and the obtained fused image represents much more informative and useful than any other input images. In this paper we have used DWT fusion technique [2.1]. Compression technique is then done on the fused images. The main purpose of compression techniques is to reduce the redundant information in an image [2]. [3]. Redundancies can be in the form of inter pixel level, Visual redundancy, Coding redundancy The proposed system was based on SVD (SINGULAR VALUE DECOMPOSITION) [4][9], where our designed system represents Truncated SVD or Reduced SVD. In Reduced SVD the reduction is done by setting first large singular values to zero and using first largest values of U and V matrix [2.2]. The obtained output is then compared from the standard method of image retrieval, the output images are also uploaded to the server using IOT (Internet Of Things). @IJRIER-All rights Reserved -2017 Page 100
Figure 1 block diagram II. BASIC MATHEMATICS 2.1. WAVELET TRANSFORM WAVELET IS REPRESENTED IN TWO FUNCTIONS SCALING FUNCTION AND WAVELET FUNCTION/ MOTHER WAVELET. WAVELET TRANSFORM [11] ARE CLASSIFIED INTO CONTINUOUS WAVELET TRANSFORM (CWT) AND DISCRETE WAVELET TRANSFORM (DWT). DWT uses filters to analyze and retrieve original signals. These filters are separated into two frequency levels low frequency and high frequency levels [5]. DWT are further classified as follows. 1) Haar Wavelets 2) Daubechies Wavelets 3) Dual tree complex Wavelets 4) Mallat Haar Wavelets can be in one scale dimensions and two scale dimensions [6]. In one scale dimensions it is divided into four sub-bands as shown in Fig 2. Haar Wavelet gives both forward transform and reverse transform, the scaling and wavelet transformation of a matrix are obtained by adding two adjacent samples divided by 2 and subtracting two adjacent samples and divided by 2 respectively [6]. Inverse of the sample is obtained by simple addition and subtraction. LL0 HL0 LL1 HL1 HL0 LH1 HH0 Input Image LH0 HH0 LH0 HH0 (1- Scale DWT) (2-Scale DWT) Figure 2 Discrete Wavelet Transform Available Online at : www.ijrier.com Page 101
2.2. REDUCED SVD In Singular Value Decomposition of a matrix of m by n is factorized into three matrices as shown below A= U * Ʃ * V' Where U, V are orthonormal matrices and Ʃ is called as Diagonal matrix. We apply SVD technique by assigning the lowest singular values which are present in the diagonal matrix to zero, by which compression on an image is occurred. The negligible discarding of the valued does not affect much on the input images [4]. In Reduced SVD we reduce the matrix by using the specified K singular values of U, V and making the largest singular values in diagonal matrix to zero. If K is chosen high then we get more information and less distortion, if we chose smaller K value we will get less information and more distortion in an image, it is if K is smaller the image is compressed much. We have to simultaneously maintain the distortion level in an image. [4]. Compression Ratio is defined as the ratio between the input image and the compressed image CR = m*n / K*(m+n+1) Where m, n are row and column of the given input matrix A and K is the singular values obtained from the diagonal matrix Ʃ. The obtained Compression ratio for different K values for SVD are shown in the below TABLE 1 TABLE 1 COMPRESION RATIO FOR SVD SNO K VALUES COMPRESSION RATIO (CR) 1 2 138.0200 2 12 23.588 3 22 12.101 4 52 5.72 5 112 2.01 6 202 1.580 7 251 1.564 8 260 1.552 9 262 1.544 10 264 1.536 The below Fig 3 represent the flow chart of the image retrieval, where the input obtained from the Sensor/Devices [8] is Video which is converted into frames, the standard conversion of video into frames are 24 frames/sec for these images we then apply Fusion and Compression technique. We compare the obtained compression values with the Standard SVD Compression ratio Table1. We then upload the output images to the main server. Available Online at : www.ijrier.com Page 102
Figure 3 Flow chart representation III. EXPERIMENT AND RESULTS STEP 1 converting the input video into frames as shown in Fig4. We get n number of image frames according to the size of the video [7]. Each image extracted from the video is referred as frames, mostly the total frames extracted per second is around 24-30 and hence called as 24FPS (Frames per Second). Figure 4 Video to image Frames STEP 2 Using Haar Wavelet and applying color map to the output of the Fusion we get the output as shown in Fig 5. Available Online at : www.ijrier.com Page 103
Figure 5 Colored Fusion Output STEP 3 Compression output using Reduced SVD according to the K singular values are shown in Fig 6. Figure 6 K Values for Reduced SVD Compression Techniques STEP 4 The obtained Compression values are represented in Table2 which can be compared with the standard SVD compression technique as shown in Table1. We could get the best compression ratio (Cr) [4]. Available Online at : www.ijrier.com Page 104
TABLE 2 CR VALUES FOR REDUCED SVD SNO K VALUES REDUCED SVD CR VALUES 1 2 127.875 2 12 21.3125 3 22 11.6250 4 52 4.9183 5 112 2.2835 6 202 1.2661 7 251 1.0189 8 260 0.6837 9 262 0.6761 10 264 0.6688 STEP 5 Uploading the output images to the server, here we have used one-drive Microsoft server where it shows the size of images uploaded as shown in Fig 7.1, Fig 7.2. Figure 7.1 Video to Frame File Figure 7.2 after applying Image Retrieval Techniques IV. CONCLUSION From the outputs and the given Table2 it can be concluded that applying DWT and Reduced SVD techniques it gives us the better results than using the standard SVD, for Big-data images we could reduce the size and also solve few problems faced in IOT applications. The proposed paper Available Online at : www.ijrier.com Page 105
can be applied in medical fields, Malls, Security purpose, it also reduces issues faced for transmitting to the main servers. REFERENCES [1] V. Krishna Sree and T. Divija Survey on Image Compression Techniques and IOT Challenges CIIT International Journal of Digital Image Processing, Vol 9, No 3, March 2017. [2] Ms. Pallavi M. Sune Prof. Vijaya K. Shandilya Amravati university Image Compression Techniques based On Wavelet and Huffman Coding ijarcsse Volume 3, Issue 4, April 2013. [3] Miss Samruddhi Kahu Ms. Reena Rahate Marvell Technologies Image Compression using Singular Value Decomposition International Journal of Advancements in Research & Technology, Volume 2, Issue 8, August-2013. [4] K.M.Aishwarya, Rachana Ramesh, Preeti.M.Sobarad, Dr. Vipula Singh Lossy Image Compression using SVD Coding Algorithm IEEE WiSPNET 2016 conference. [5] Sandeep Kaur Gaganpreet Kaur Dr. Dheerendra Singh A Review: Various Wavelet Based Image Compression Techniques IJSR Volume : 2 Issue : 5 May 2013. [6] Andrea Gavlasov a, Aleˇs Proch azka, and Martina Mudrov a WAVELET BASED IMAGE SEGMENTATION Institute of Chemical Technology, Department of Computing and Control Engineering. [7] Punith Kumar M B 1, Dr. P.S. Puttaswamy2 VIDEO TO FRAME CONVERSION OF TV NEWS VIDEO BY USING MATLAB IJARSE. [8] M.Pradeep Implementation of Image Fusion algorithm using MATLAB (LAPLACIAN PYRAMID) 2013 IEEE. [9] Deepika Sharma Pawanesh Abrol Experimental Analysis of Digital Image Retrieval Using SVD 2014 International Conference on Computing for Sustainable Global Development. [10] Moeen Hassanalieragh, Alex Page, Tolga Soyata, Gaurav Sharma, Mehmet Aktas, Gonzalo Mateos Burak Kantarci, Silvana Andreescu Health Monitoring and Management Using Internet-of-Things (IoT) Sensing with Cloud-based Processing: Opportunities and Challenges 2015 IEEE International Conference on Services Computing. [11] Nisha Gawari, Dr. Lalitha.Y.S Comparative Analysis of PCA, DCT & DWT based Image Fusion Techniques International Journal of Emerging Research in Management &Technology Volume-3, Issue-5. Available Online at : www.ijrier.com Page 106