The following images are for the Missouri-B model. The method used was
linear prediction, with 3 ancestors, and values were quantized BEFORE linear
prediction. (A) is the original Missouri-B model, while (B)-(F) are its
8-16-bit counterparts. The original Missouri-B model is 154,871 bytes,
while the respective compression ratios are below each image.

CR = 52.32 : 1 size = 2,960 bytes
A
B

CR = 40.16 : 1 size = 3,856 bytes
CR = 31.79 : 1 size = 4,872 bytes
C
D

CR = 28.12 : 1 size = 5,507 bytes
CR = 28.47 : 1 size = 5,440 bytes
E
F
If you have comments or suggestions, email me at scottm@meru.cecs.missouri.edu