
176
5 Compression
We conclude this section by briefly describing several types of visual impair-
ments in a picture that are unique to predictive compression. These include
slope overload noise, edge busyness, granularity noise, and contouring noise
[5,
59].
At sharp transitions in a picture, the error
e(m,n)
is usually very large.
If it is larger than the greatest quantizer step, the system is said to be in "slope
overload." The result is that an abrupt transition may appear smeared, since
when the system gets into slope overload, it may take several samples to recover.
A nonuniform quantizer may be designed with large outer steps to force