Low-memory-usage and high-speed image wavelet transform

Abstract

Discrete wavelet transform (DWT) is being broadly used in processing digital image, but the large requirement of memory and the time delay limit the DWT usage. For instance, when the image is processed in the space, the realtime processing, low power consumption, reduced complexity and low memory consumption are required. Lifting-based DWT is implemented to lessen complexity. And then the coefficients of the lifting filters are turned to be binary and the filters are therefore implemented efficiently without using any multiplier. In this way the frequency of DWT can be improved and the architecture is simplified. When 2-dimensional DWT is carried out, the line-based wavelet transform is able to save memory in a larger sense. The architecture of line-based wavelet can also be executed in a parallel way. Both horizontal transform and vertical transform can be executed at the same time. This paper takes 9/7 wavelet as an example and compares with others its architecture and performance. The advantages of this design include real-time performance improvement, memory reduction and architecture simplification.