DEVELOPMENT OF A NON-ITERATIVE METHOD OF MATRIX FACTORIZATION FOR IMAGE COMPRESSION AND TRANSMISSION OF THE FACTORIZED COEFFICIENTS USING DIFFERENTIAL SIGNALING BASED BPSK SYSTEM
N. Nithiyanandam, Bonafide Certificate · 2014
Data compression is an essential requirement of digital information storage, transmission and retrieval. Digital images form a significant part in multimedia systems. This thesis is a report on the development of a non-iterative based positive matrix factorization technique for image data compression, along with a differential signaling based BPSK system for data transmission. Although positive matrix factorization is a known technique for image compression, the technique is based on iterative method for factorization. In this research, a very useful non-iterative method for positive matrix factorization has been developed using AH-DWT (Asymmetric Haar – Discrete Wavelet Transform) and tested using medical, electron microscopic and satellite images. This 2D matrix factorization technique has also been extended to 3D matrix factorization and applied for hyper spectral image compression. Subsequently a differential signaling based BPSK system has been developed for the transmission of factorized coefficients and tested for its performance in AWGN (Additive White Gaussian Noise) channel and phase error causing Rayleigh and Rician channels. A detailed mathematical analysis has been made on the performance of differential signaling based BPSK in transmission channels affected by Gaussian noise and phase noise separately. The developed transmission system has also been tested using medical image data. Admittedly, a number of efficient channel coding schemes exist, but with the requirement of overhead bits. The differential signaling based BPSK transmission has an inherent ability for partial detection of errors without any overhead bits. With suitable channel coding schemes at the cost of overhead bits, the proposed system will perform better.