Game Development Reference
In-Depth Information
Fig. 7.4 One-dimensional
visualization of pixels in a
filtering block unit
Table 7.1 The relationship between BS and pixel variations
Pixel variations
BS
fS equal to 6, p 0 equal p 1 and q 0 equal to q 1
4
fS equal to 6, p 0 equal p 1 and q 0 equal to q 1 are not simultaneous; or
3
fS equal to 5, p 0 equal p 1 and q 0 equal to q 1
fS equal to 6, p 0 equal p 1 and q 0 equal to q 1 are not simultaneous; or
2
fS equal to 4, fL equal 2
fS equal to 3, absolute value of ( p 1 q 1 ) smaller than ʲ
1
Others
0
assigned according to Table 7.1 , which illustrates the relationship between BS and
pixel variations.
1
< |
p 0
q 0 | <ʱ.
(7.1)
Here, the notations p 0 ,
p 1 ,
p 2 ,
q 0 ,
q 1 ,
q 2 are the sample values before filtering
and P 0 ,
Q 2 are the filtered sample values. When the constraint in
Eq. ( 7.1 ) is not satisfied, BS value is directly assigned as 0. The variables fL , fR ,
and fS in Table 7.1 are derived according Eqs. ( 7.2 )-( 7.4 ).
P 1 ,
P 2 ,
Q 0 ,
Q 1 ,
fL
= ( | p 0
p 1 |
?2
:
0
) + ( | p 0
p 2 |
?1
:
0
),
(7.2)
fR
= ( |
q 0
q 1 |
?2
:
0
) + ( |
q 0
q 2 |
?1
:
0
),
(7.3)
fS
=
fL
+
fR
.
(7.4)
, are adaptive thresholds based on the average quanti-
zation parameters (QP) of two adjacent blocks, and their relationship with QP is illus-
trated in Fig. 7.5 . When the QP is small, blocking artifact is weak and the boundary
discontinuities are likely caused by image edges, which should be preserved. Then
the two thresholds should be small. When QP is large, blocking artifacts are more
The two parameters,
ʱ
and
ʲ
 
Search WWH ::




Custom Search