We propose a novel lossless audio compression scheme, which combines stereo preprocessing with stereo prediction. We show that such a scheme provides improved asymmetrical compression at almost no complexity increase for decoder (compared with stereo prediction alone), or the same compression for lower decoder complexity. The stage of stereo prediction is preceded by a rotation-like channel transformation, which improves compression by requiring smaller inter-channel optimal prediction orders and by obtaining smaller magnitudes for prediction coefficients For the OptimFROG-AS (asymmetric) lossless audio compressor using stereo prediction with orders 8/4 we obtained, on an audio corpus (in CD-Audio format) of size 51.6 GB, compression improvements up to 5.10%, on average 0.23%.
展开▼