An Optimal Wavelet Filter-Based Method that Reduces Execution Time for Unstructured Image Compression Algorithms

The dramatic increase in the computational speed provided by state-of-the-art high-end DSP’s offers the possibilities of processing large high-resolution images in real time. The authors have experience in high-fidelity lossy compression of high-resolution color medical images (up to 2048x4096 pixels, eight bits per pixel) using TI’s TMS320C6201 DSP’s. The amount of on-chip memory in C’6000 DSP cores is insufficient for processing such images using internal memory only. Allocating the data to the external memory can increase the execution time significantly. The use of direct memory access channels (DMA) does not lead to a substantial decrease of the execution time for unstructured algorithms, i.e. the algorithms in which the data processing order is highly dependent on the results of the previous steps in the algorithm. We have developed and tested a technique, which decreases execution time for such algorithms by applying them to only one subband of a signal dependent wavelet transform.