Modified Davis and Peet 
Spatial Filtering 
'IBIS' Spatial 
Filterinq 
Mod. Davis and Peet fol 
lowed by 'IBIS' Filter 
Original 
^After Sim- 
2 
Pixels Un 
changed by 
1 After Sim- 
2 
Pixels Un 
changed by 
^After Sim- 
2 
Pixels Un 
changed by 
CLASS 
Histoqram 
piification 
Simpl ific. 
pi i fication 
Simplific. 
piification 
Simplific. 
M4G 
1.9 
0.9 
35.5 
1.0 
31.2 
1.0 
35.5 
M4P 
17.2 
14.8 
57.8 
20.1 
58.0 
17.2 
57.9 
M3G 
18.9 
18.0 
68.9 
18.5 
56.4 
18.8 
65.8 
M3P 
31 .4 
37.8 
86.7 
42.9 
82.7 
42.9 
84.6 
M3G/4P 
9.0 
8.0 
62.0 
4.9 
33.0 
6.7 
50.4 
Hardwd 
13.9 
16.9 
81 .7 
11.5 
51 .4 
13.4 
59.5 
Barren 
0.3 
0.3 
25.0 
0.0 
0.0 
0.0 
0.0 
Sparse 
2.7 
1 .1 
27.5 
0.0 
0.0 
0.4 
8.7 
Brush 
3.6 
2.3 
41 .0 
1 .1 
17.1 
1 .2 
20.8 
Unclas 
1 .1 
0.0 
0.0 
0.0 
0.0 
0.0 
0.0 
Percentage of total classified image. 
2 
Percentage of each class category. 
TABLE 1 -- Effects of post-processing (simplification) on a classified forest image. The histogram change 
resulting from three spatial filtering techniques can be compared against the original histogram. The per 
centage of original classified pixels unchanged by each simplification process are also shown. The Modified 
Davis and Peet technique employed individual class conversion weights. Classes are arranged in the table to 
show the order of the 'IBIS' filter technique's hierarchical class conversion weights structure (weighted 
highest to lowest). The Modified Davis and Peet method best preserves the original histogram and pixel spa 
tial structure, but permits diagonal connections. The 'IBIS' filtering method removes diagonal connections 
but experiences greater histogram and spatial change. Use of the Modified Davis and Peet approach with sub 
sequent application of the 'IBIS' filtering approach produces a better product than the 'IBIS' approach alone