Hybrid DCT_DFT_Wavelet Transform _HWT_ by pptfiles


									Course: EE 800 Name: Samia Sharmin Shimu (11072595) Md. Ashraful Islam (11072611)

Project Name: Hybrid Architecture of DCT/DFT/Wavelet Transform (HWT)
Objective: Hardware implementation of Hybrid Discrete Cosine Transform (DCT)/
Discrete Fourier Transform (DFT)/Wavelet Transform (HWT) in single chip.

1. Hardware saving: The hardware requirement for this hybrid chip is less than the total hardware requirement if DCT, DFT and HWT are implemented separately. 2. Less Power consumption

Discrete cosine transform (DCT), particularly the DCT-II is used in signal and image processing, video data compression standards such as JPEG, MPEG1, MPEG2, and H.26X. The Discrete Fourier Transform (DFT) is also a popular transformation for signal processing and Communication. Furthermore, the discrete wavelet transform based on Haar matrix (HWT) is also very useful in image compression and JPEG-2000 standard.

In our project we will implement DCT/DFT/Haar wavelet in single hybrid architecture. The Algorithm of combining DCT/DFT/HWT is obtained from the paper, Zhu Chen, Moon Ho Lee “On Fast Hybrid Source Coding Design”, 2007 International Symposium on Information Technology Convergence. We rearrange the DCT, DFT and Haar wavelet matrix coefficient in similar structure pattern to share the common block diagrams. As DCT is based on cosine transform and DFT is a combination of cosine and sine transform they have the similar cosine blocks that can be shared. Based on this principle DCT and DFT filters have been hybridized and later Haar matrix has been decomposed to similar recursive computational pattern as the simplified coefficient used in Haar matrix is also present in DCT. Different combination of matrix algorithm has been use to decompose the classical DCT, DFT and Haar matrix to find out the common blocks which will be shared by the three filters in our architecture. This way it will require less hardware for our new propose hybrid architecture for the DCT, DFT & Haar filters rather than using three different filters. In this project we will build the hardware in one dimensional platform and all the filters will be mutually exclusive.

To top