International Journal of Computer Applications (0975 – 8887) Volume 68– No.21, April 2013 22 Advance Technique for Feature Extraction and Image Compression Arvind Kourav Electronics & Communication Engineering Research Scholar D.K.N.M.U. Niwai Rajasthan India Prashant Singh, PhD. Computer Science & Engineering ISC Software Pvt. Ltd., Bhopal India ABSTRACT For image processing, it is very necessary that the selection of transform. In this paper, a comparative analysis of curve let transform with other transform for image processing .In this we proposed the applications of curve let transform in the field of image Compression ,phase recognition and feature extraction. For higher compression with quality reconstruction .The Wavelets gave a different aspect to the compression. Curvelet Transform gives better results in terms of PSNR. Face recognition is very important for many applications such as: video surveillance, criminal investigations and forensic applications, secure electronic banking, mobile phones, credit cards, secure access to buildings . The curve let transform is a multi scale directional transform, which allows an almost optimal non adaptive sparse representation of objects with edges. Curve let have also proven useful in diverse fields beyond the traditional image processing application, Curvelet transform improve recognition accuracy with featature extraction extraction algorithms PCA, LDA,ICA and NMF. Keywords Image processing, Image Compression, Feature Extraction, Curvelet transform, Wavelet Transform. 1. INTRODUCTION In the field of Image processing different types of method can be use for the analysis of image in different field such as De noising , Compression ,Face recognition ,Bio medical application etc. These analysis of image is based on the different types of transform is Fourier transform, Wavelet Transform, Curve let transform , Now a day Wavelet and Curvelet transform is used in all field of image processing. 1.1 DIGITAL IMAGE PROCESSING SYSTEM A typical digital image processing system consists of image segmentation, feature extraction, pattern recognition, thresholding and error classification. Image processing aims at extracting the necessary information from the image. The image needs to be reduced to certain defining characteristics and the analysis of these characteristics gives the relevant information. 1.2 CURVELET TRANSFORM To overcome the draw back of Wavelet Transform .Curve let Transform is developed, Curve let Transform is very effective modal that not only consider a multi scale Time Frequency local portion but also make use of the direction of features. It was developed by Candes and Donoho in 1999,there are two types of Curvelet transform is unequally spaced Fast Fourier Transform and wrapping based fast Curve let Transform, in curve let transform the width and length are related by the relation Width~ Length 2 that is known as parabolic or anisotropic scaling . Moreover, frame elements in curve let indexed by scale, location and orientation parameters in contrast to wavelets where elements have only scale and location. This transform can we used for both continuous and digital domain ,In this angle polar wedges or angle trapezoid window is used in frequency domain .Initially construction of Curve let transform is redesigned as fast discrete Curve let transform (FDCT) by Candes in 2006.This is second generation curve let transform is meant to be simpler to understand and use . DCT can be implemented by wrapping based fast discrete curve let at a given scale. At a given scale and orientation both the image and the curve let are transformed into Fourier domain. The product of curve let and the image are obtained in the fourier domain. Inverse FFT is applied to the above product to obtain a set of curve let coefficients. In order to perform IFT the trapezoidal wedge thus obtained from the frequency response of a curve let is wrapped into a rectangular support. The spectrum inside the wedge is tilted periodically. Thus the rectangular region collects the wedge’s fragmented portions by periodic tilting. The paper is structured as tracks: the Image Compression are discussed briefly in section II; section III describes feature extraction methods .The result analysis is discussed in section IV and section V forms the conclusion. 2. IMAGE COMPRESSION Basic steps of the image compression is shown in fig. 1. All these steps are invertible, therefore lossless, except for the Quantize step. Quantizing is the process of reduction of the precision of the floating point values of the curvelet transform. Fig 1: Image Compression Block Diagram Mostly, scalar quantization is used for quantization approach [16.18]. But, is significantly affects the bitrate every time a run of zeros is broken. All nonzero coefficients, independent of size, have the same negative impact in the run length encoder. Small coefficients enhance the image quality to a great extent and they still mess up the Huffman encoding as much as the big coefficients. Figure 2 shows the utilization of a proposed quantizer rather than a scalar quantization by which it is possible to block the smallest coefficients [19]. Longer runs of zeros are now attained, and this improves the performance of the Huffman encoder. The performance of the entire system is improved in terms of psnr/bpp Curvelet Coefficient Quantiz ation process Huffman Coding Input Image