A generic 2D discrete wavelet transform on FPGA

Abstract

Advancements in the telecommunication environment and the trend of having internet over smartphones and tablets which have limited storage and bandwidth, have led to an increase in the need for better image and video compression. In this regard, new compression methods are being introduced. The 2D Discrete Wavelet Transform has proven to be one of the best techniques to achieve high quality compression ratios with very low image distortion. The objective of this thesis was to design and implement a Generic 2D Discrete Wavelet Transform on an FPGA that can be used as a component in other projects. This had to include the possibility of adjusting the number of significant bits and the number of levels of transform that needs to be performed on the image. A comprehensive study of the Discrete Wavelet Transform (DWT) on images was presented. It was established that the DWT could be implemented using two different methods - convolution and lifting. Thus both of these techniques were realised using MATLAB. The outputs were then investigated and the results were comparable. Due to the fact that the lifting algorithm is known to be more efficient for hardware implementation, it was implemented using VHDL. The implemented wavelet transform block can produce up to three levels of transforms. When the entire design and implementation were successfully simulated and tested on an FPGA with a 100 MHz clock, the code was then optimised in order to become more efficient. With this optimisation, the performance of the whole system was increased by a speed factor of 7 .5 times when compared to the original unoptimised code