2-D DWT system architecture for image compression

Abstract

Since the introduction of wavelet transform in early 1980s, it has contributed significantly in multiple areas, such as image processing and compression, radar signal analysis, numerical analysis, biomedical signal processing, medical imaging and digital signal processing. The key advantage of wavelet analysis is the extra time and frequency information compared to other transforms. However, the discrete wavelet transform (DWT) requires very large memory requirement and is computationally intensive, especially for 2-D transform. Typically, it has quadratic computational complexity. In this project, we propose a fully dedicated processor which specialized in 2-D DWT. This architecture aims to achieve improvements on throughput, scalability and flexibility compared to other prior architectures. This architecture requires significantly less computational resources and internal memory. The proposed architecture can achieve threoritical throughput of 138fps for a 2048x1566 video processing. The DWT system has been designed for scalability by supporting up to 8 parallel DWT engines and each DWT engine can work independently. The DWT system architecture is very flexible and the performance can be scaled by increasing or reducing the DWT engines, according to different application needs. Furthermore, this architecture has been designed with the consideration of integration into existing Advanced Microcontroller Bus Architecture (AMBA) systems in future.