الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 116 |  |  |
| | | from | 116 | | |
المستخلص
Multiresolution analysis achieved by wavelet transform is used in many applications for coding and compression of images and video streams. It is implemented using multirate filter banks. In this thesis we compare hardware efficiency and compression performance of some of the important filter-bank structures and apply them for the implementation of the biorthogonal wavelet transform used for wavelet image coding.
One of the famous wavelets is the Daubechies 9/7 biorthogonal wavelet transform. It is specified in JPEG 2000 image coding standard, and it will be our concern throughout this thesis.
The comparison includes convolution-based structures, as the direct-form, cascade, and polyphase structures, and the lifting-based structure. Algorithms for all four structures are simulated and the resulting magnitude responses are compared at different finite-word lengths to record the finite-word-length effect. In addition, such algorithms are used for compression performance comparison in a JPEG 2000 engine. Moreover, functional block modeling for all structures is also simulated and verified. Finally, hardware efficiency is compared.
Although a lifting-based filter bank structure requires half the number of mathematical operations for realizing a biorthogonal wavelet filter bank, direct-form and polyphase structures have a better finite-word-length performance. Moreover, operations in a polyphase filter bank require processing at a half the rate required by lifting-based filter banks; an issue which was neglected by the reviewed literature.