Switching of Wavelet Transforms by Neural Network for Image Compression

Switching of Wavelet Transforms by Neural Network for Image Compression

Houda Chakib (Physics Sciences Department, Sultan Moulay Slimane University, Beni Mellal, Morocco), Brahim Minaoui (Physics Sciences Department, Sultan Moulay Slimane University, Beni Mellal, Morocco), Abderrahim Salhi (Physics Sciences Department, Sultan Moulay Slimane University, Beni Mellal, Morocco) and Imad Badi (Computer Sciences Department, Sultan Moulay Slimane University, Beni Mellal, Morocco)
Copyright: © 2018 |Pages: 14
DOI: 10.4018/JECO.2018010104
OnDemand PDF Download:
No Current Special Offers


Nowadays, digital images compression requires more and more significant attention of researchers. Even when high data rates are available, image compression is necessary in order to reduce the memory used, as well the transmission cost. An ideal image compression system must yield high-quality compressed image with high compression ratio. In this article, a neural network is implemented for image compression using the feature of wavelet transform. The idea is that a back-propagation neural network can be trained to relate the image contents to its ideal compression method between two different wavelet transforms: orthogonal (Haar) and biorthogonal (bior4.4).
Article Preview

1. Introduction

In the past few years, there has been a fast development of the computer applications which caused a big increase of the use of digital data mainly in the domain of multimedia, medical imagery, satellite transmission, games, etc., and this data need in some point to be compressed in order to reduce the storage space and the transmission cost, consequently many techniques of compression such wavelet transforms are been rapidly developed to handle and store data in efficient way (Ahmadi, Javaid & Salari, 2015; Rabbani & Jones, 1991).

Wavelet transforms are lossy powerful methods that compress images at higher compression ratio; they are a time-frequency representation that takes account both the time and the frequency of the signal to analyze (Rabbani & Jones, 1991). They have been used with great success in a wide range of applications. For many years, a number of researchers have been interested in the three important processing stages of the wavelet compression technology: pixels transform, vector quantization and entropy coding (Krishnanaik, Someswar, Purushotham & Rajaiah, 2013) and the majority of the articles they published have shown the efficiency of the technique in compressing images at high compression ratios with low image distortion (Antonini, Barlaud, Mathieu & Daubechies, 1992; Said & Pearlman, 1996). As a result of all this research, several wavelet families have emerged (Gandhi, Panigrahi & Anand, 2011) and all these techniques have the same purpose which is to yield high compression ratio while maintaining high-quality images.

The implementation of artificial neural networks in image processing application has increased since they have been introduced in the signal processing few years ago (Zang & Beneveniste, 1992; Jiang, 1999). Moreover, a lot of research has focused on the combination of image compression techniques and the use of the neural network with great success. This combination has proved to be a valuable tool for image processing (Dimililer, 2013; Dimililer & Khashman, 2008; Alexandridis & Zaprani, 2013).

Despite the fact that the wavelet library contains a large number of wavelets, many researchers have proved that the choice of the best wavelet has significant impact on the quality of compression (Gandhi, Panigrahi & Anand, 2011). Therefore, the work presented in this paper deals with the selection of the most suitable wavelet function for compressing a particular image between two wavelets transforms Haar and Bior4.4 by training a back-propagation neural network to relate a grayscale image to its ideal compression system just by learning the nonlinear relationship between the pixel intensities of the image.

This paper is organized as follow: section 2 is shared in three parts: the first part presents a brief theory of the wavelet transform process, the second part focuses on the theoretical concept of the feed-forward artificial neural network the last part presents the back-propagation based on the scale conjugate gradient algorithm. In section 3, the methodology of the experiment is presented followed by the results and discussions in the section 4. Finally, section 5 demonstrates conclusion and directions for future works.

Complete Article List

Search this Journal:
Volume 20: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 19: 4 Issues (2021)
Volume 18: 4 Issues (2020)
Volume 17: 4 Issues (2019)
Volume 16: 4 Issues (2018)
Volume 15: 4 Issues (2017)
Volume 14: 4 Issues (2016)
Volume 13: 4 Issues (2015)
Volume 12: 4 Issues (2014)
Volume 11: 4 Issues (2013)
Volume 10: 4 Issues (2012)
Volume 9: 4 Issues (2011)
Volume 8: 4 Issues (2010)
Volume 7: 4 Issues (2009)
Volume 6: 4 Issues (2008)
Volume 5: 4 Issues (2007)
Volume 4: 4 Issues (2006)
Volume 3: 4 Issues (2005)
Volume 2: 4 Issues (2004)
Volume 1: 4 Issues (2003)
View Complete Journal Contents Listing