This paper sheds light on the recent least-square (LS)-based adaptive predi
ction schemes for lossless compression of natural images. Our analysis show
s that the superiority of the LS-based adaptation is due to its edge-direct
ed property, which enables the predictor to adapt reasonably well from smoo
th regions to edge areas. Recognizing that LS-based adaptation improves the
prediction mainly around the edge areas, we propose a novel approach to re
duce its computational complexity with negligible performance sacrifice. Th
e lossless image coder built upon the new prediction scheme has achieved no
ticeably better performance than the state of-the-art coder CALIC with mode
rately increased computational complexity.