EURASIP Journal on Applied Signal Processing 
Volume 2002 (2002), Issue 9, Pages 981-989
doi:10.1155/S1110865702204126

Wavelet Kernels on a DSP: A Comparison between Lifting and Filter Banks for Image Coding

Stefano Gnavi, Barbara Penna, Marco Grangetto, Enrico Magli, and Gabriella Olmo

 CERCOM, Center for Multimedia Radio Communications, Dipartimento di Elettronica, Politecnico di Torino, Corso Duca degli Abruzzi 24, Torino 10129, Italy
Received 30 August 2001; Revised 30 April 2002

Abstract

We develop wavelet engines on a digital signal processors (DSP) platform, the target application being image and intraframe video compression by means of the forthcoming JPEG2000 and Motion-JPEG2000 standards. We describe two implementations, based on the lifting scheme and the filter bank scheme, respectively, and we present experimental results on code profiling. In particular, we address the following problems: (1) evaluating the execution speed of a wavelet engine on a modern DSP; (2) comparing the actual execution speed of the lifting scheme and the filter bank scheme with the theoretical results; (3) using the on-board direct memory access (DMA) to possibly optimize the execution speed. The results allow to assess the performance of a modern DSP in the image coding task, as well as to compare the lifting and filter bank performance in a realistic application scenario. Finally, guidelines for optimizing the code efficiency are provided by investigating the possible use of the on-board DMA.