1. INTRODUCTION
It is often useful to implement compression algorithms that account for the ultimate classification task associated with the decoded imagery, such as in detecting biological abnormalities in compressed medical images and in compressing aerial imagery for remote-sensing applications. The goal is to compress the image efficiently while accounting for the fact that the decompressed image will be employed in a classification task. The overall encoding scheme is shown in Fig. 1. We here focus on wavelet-based image compression algorithms, since such now represent the state of the art and are used in practical algorithms.