Please use this identifier to cite or link to this item:
Title: Ondelettes et techniques de compression d’images numérique
Authors: ZITOUNI, Athmane
Keywords: Image Compression
entropy coding
compression ratio
compression ratio
Issue Date: 2013
Abstract: This thesis is devoted to the study of transforms applied in the literature (wavelet transform directional, fourrier ...) in the context of digital image compression. We also address the study of main coding methods used in image compression as (Shannon Fano coding, Huffman jepg2000, hierarchical coders ...). We present the influence of new mathematical properties provided by the wavelet theory in the field of hierarchical coding for application to digital image compression. We show by theoretical analysis that the multi-resolution decomposition of the image, the practical contribution of the wavelet theory is needed. Therefore, we study the techniques of digital image compression. The advantage of multi-resolution analysis results in its decomposition into pyramids. The use of hierarchical coders is based on the concept of zero tree (zerotree). We propose a new approach for image compression based on the basic principle of the SPIHT algorithm. We note that our new approach denoted MSPIHT (Modified SPIHT) is to minimize the bit code after quantization. We are looking for it to encode several coefficients using a single bit which is sufficient for our approach as for the basic method (SPIHT) it is not. The results obtained by this new approach we propose based on the following metric: • PSNR • Compression ratio Its better than the results obtained by the basic method (SPIHT Amir SAID) [36]. Our contribution is best especially for medium and high speeds without affecting the computation time. Finally, our results are comparable to those obtained by the algorithms SPIHT, EZW and JPEG 2000.
Appears in Collections:Département de Génie Electrique

Files in This Item:
File Description SizeFormat 
ondelette_et_technique_de_compression_d_image_numerique.pdf2,6 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.