Reduced complexity VLSI implementation of discrete wavelet transform for image and video compression

In the recent years, image and video coding techniques have gradually moved from the DCT-based to the DWT-based coding systems. At the same bit rate, the latter provides better coding efficiency. The success is testified by the fact that DWT was adopted by the FBI Fingerprint, the JPEG 2000 and the MPEG-4 standards.; DWT, however, is an intensive computational process. The complexity of a wavelet-based coding system is several times higher than that of the DCT-based approach. Most of mobile multimedia devices have limited resources of CPU, memory and power. There is a need to investigate and develop computationally efficient VLSI architectures to implement DWT for these applications. To achieve this goal, this dissertation focuses on the theory, design, development and implementation of efficient VLSI architectures for DWT.; Hardware implementation of DWT has some basic problems. First, DWT, in general, requires floating-point operations. Floating-point DWT involves more computations, has long latency, and requires more memory for storing intermediate results. Another hurdle for hardware implementation of DWT relates to the hardware cost and the delay of multipliers. In order to optimize the performance and to reduce hardware complexity of DWT, both floating point and multiplication operations should be minimized or eliminated.; In this research, we present an efficient method to reduce complexity for DWT implementation. The proposed method is based on the design and implementation of integer wavelet filters. From the VLSI design viewpoint, integer wavelet filters have many attractive features. They can be realized in hardware with very small footprints. In addition, power consumption is significantly reduced since floating-point arithmetic units and multipliers are eliminated. Moreover, the throughput of DWT system is improved. Furthermore, the integer wavelet transform provides a natural way for building unified lossy/lossless image compression systems.; In this dissertation, we present the integer wavelets designed for lossy compression in JPEG 2000 standard. We then introduce a general framework to design integer wavelets. Several low power, high performance VLSI architectures for convolution and lifting schemes are presented. In this study, we show that both low power and high performance hardware implementations of DWT can be achieved at the algorithm-architecture level.