Lossless Coding of Multiband Images Using Interband Data Correlation and ErrorFeedback Prediction Scheme

摘要

In this paper, a multiband lossless compression system exploiting inter-band data correlation is presented. We develop an adaptive prediction scheme that can dynamically switch among smooth, intra and intra-band prediction modes subject to the data correlation. Specifically, we use context information to determine the inter-band correlation between neighboring bands, and then borrow the wisdom of optimal prediction formula from the previous band if its context matches the current one. Prediction error from the previous band is feedback for bias cancellation. Experimental results show that the proposed system delivers the best coding performance among various multi-band image formats. The bit-rate saving over other state-of-the-art techniques can be up to 16%.