www.ijecs.in International Journal Of Engineering And Computer Science ISSN:2319-7242 Volume 2 Issue 8 August, 2013 Page No. 2468-2475 V.Krishnanaik, IJECS Volume 2 Issue 8 August, 2013 Page No.2468-2475 Page 2468 IMPLEMENTATION OF WAVELET TRANSFORM, DPCM AND NEURAL NETWORK FOR IMAGE COMPRESSION V.Krishnanaik 1 Dr.G.Manoj Someswar 2 K.Purushotham 3 A.Rajaiah 4 Abstract: Images have large data capacity. For storage and transmission of images, high efficiency image compression methods are under wide attention. In this paper we implemented a wavelet transform, DPCM and neural network model for image compression which combines the advantage of wavelet transform and neural network. Images are decomposed using Haar wavelet filters into a set of sub bands with different resolution corresponding to different frequency bands. Scalar quantization and Huffman coding schemes are used for different sub bands based on their statistical properties. The coefficients in low frequency band are compressed by Differential Pulse Code Modulation (DPCM) and the coefficients in higher frequency bands are compressed using neural network. Using this scheme we can achieve satisfactory reconstructed images with increased bit rate, large compression ratios and PSNR. Keywords: Efficiency, subband, Huffman coding, DPCM, PSNR, Haar wavelets 1. INTRODUTION Image compression is a key technology in the development of various multi-media computer services and telecommunication applications such as video conferencing, interactive education and numerous other areas. Image compression techniques aim at removing (or minimizing) redundancy in data, yet maintains acceptable image reconstruction. A series of standards including JPEG, MPEG and H.261 for image and video compression have been completed. At present, the main core of image compression technology consists of three important processing stages: pixel transforms, vector quantization and entropy coding. The design of pixel transforms is to convert the input image into another space where image can be represented by uncorrelated coefficients or frequency bands. Therefore, only those main frequency bands or principal components are further processed achieve image compression such as DCT, wavelets, etc. Vector quantization rounds up the values of transformed coefficients into clusters where points within a cluster are closer to each other than to vectors belonging to different clusters. Entropy coding is a form of lossless data compression in which statistical information of input data considered to reduce the redundancy. Typical algorithms are arithmetic coding, Huffman coding and run-length coding etc. The use of sub band decomposition in data compression and coding has a long history. Sub band coding was first proposed by Crochiereet al. for medium bandwidth waveform coding of speech signals. This method decomposed the signal into different frequency bands using a bank of quadrature mirror filters(QMF‘s). Each sub band was subsequently encoded using differential pulse code modulation(DPCM). A varying bit assignment strategy was also used to allocate the bit rate for each sub band according to its statistical properties. Woods and O‘Neil extended sub band decomposition to two-dimensional (2-D) signals and proposed a method for QMF design that eliminates possible aliasing error due to non ideal sub band filters. Recent advances in signal processing tools such as wavelets opened up a new horizon in sub band image coding. Studies in wavelets showed that the wavelet transform exhibits the orientation and frequency selectivity of images. Neural networks approaches used for data processing seem to be very efficient, this is mainly due to their structures which offers parallel processing of data and, training process makes the network suitable for various kind of data. Sonhera et al have used a two layered neural network with the number of units in the input and output layers the same, and the number of hidden units smaller. The network is trained to perform the identity mapping and the compressed image is the output of the hidden layer. Arozullah et al presented a hierarchical neural network for image compression where the image is compressed in the first step with a given compression ratio; then the compressed image is itself compressed using another neural network. Hussan et al proposed a dynamically constructed 1 Asst. Professor, Department of Electrical & Computer Engineering, College of Engineering & Tech, Aksum University,Axsum. E-Mail: krishnanaik.ece@gmail.com, 2 Professor in CSED, Anwarul- uloom College of Engineering and Technology, Hyderabad. India. E-mail: manojgelli@gmail.com 3 Asst. Professor, Department of Electrical & Computer Engineering, College of Engineering & Tech, Jigjiga University, Jigjiga, Ethiopia, North East Africa, Ethiopia. E-Mail: krishnanaik.ece@gmail.com, 4 Assoc.Professor, Department of Electronics & Communications Engineering, Joginpally B.R Engineering college, Hyderabad, Andrapradesh, India. E-Mail: rajua1999@gmail.com,