Improved compression of medical images by classification and wavelet filter assignment
N. Patuck · 2002
Current state-of-the-art image coding techniques rely on the wavelet transform to compact the energy of an image before coding. However, all such coders rely on a single wavelet basis to perform the transform. This is inefficient as a single basis may not be entirely suited for any given image. In this paper, we describe a technique to select the most appropriate basis by classifying an image based on certain feature measures and performing a cluster analysis on these measures. After classification, the most appropriate basis is selected for the transform, resulting in a coding gain varying from 0.25 to 0.1 dB. (4 pages)