An efficient implementation of one-dimensional discrete wavelet transform algorithms for GPU architectures

Abstract In this paper, the authors present several self-developed implementation variants of Discrete Wavelet Transform (DWT) computation algorithms and compare their execution times against commonly approved ones for representative modern Graphics Processing Units (GPUs) architectures. The proposed solutions avoid time-consuming modulo divisions conditional instructions used DWT filters wrapping by proper expansion DWTs input data vectors. main goal research is to improve popular GPU architectures while retaining code’s clarity simplicity. relations between time improvements GPUs are also compared with counterparts traditional sequential processors. experimental study shows that implementations, in case parallel realization on GPUs, characterized very simple kernel code high performance.