A lifting based system for optimal compression and classification in the JPEG2000 framework

In this paper, we propose a novel design for a lifting based system that achieves the optimal trade off between compression and classification performances. In addition, it can also achieve superior compression performance compared to existing wavelet kernels. The proposed system is based on bi-orthogonal filters and can operate in the JPEG2000 framework. Typically, in our proposed system the trade off point between compression and classification is automatically determined by the system. However, the user can also fine-tune the relative performance if required using two adjustors (one for compression and one for classification). Extensive simulations have been performed to demonstrate the efficiency of our system. Our simulation result shows that a 99% classification performance can be achieved at a comparable reconstruction quality whilst a superior compression performance over the JPEG2000 standard can be achieved at a classification rate of 70%.