398 
phase of this research effort is the interfacing of FOCIS-derived com 
puter data with the U.S. Forest Service's computerized Wildland Re 
source Information System (WRIS). However, the data formats of the two 
systems are incompatible. POCIS produces data in raster-image format 
which is stored in a matrix-like array, and WRIS requires vector- 
polygon data in which polygons are formed by the linking of nodal 
points. A raster-to-vector conversion technique has been developed to 
facilitate the merging of FOCIS products into WRIS format. 
Post-Processing Techniques. The primary product produced by FOCIS is a 
timber stand or stratum map which is the result of two independent 
steps. In the first step, regional forest types are predicted iron 
field calibrated ecological models which require the use of digital 
terrain data. The second step involves classifying Landsat and texture 
data into tree height and density classes. While spatial contiguity 
is an irrportant criterion for assigning labels to individual timber 
classes, no spatial contraints are imposed on the Landsat based clas 
sification process. As a result, the final classification can exhibit 
high spatial variance ("noise", "speckle"). Several approaches de 
signed to reduce high spatial variance and meet the ten acre timber 
stand size required by the USFS have been investigated. These tech 
niques include: 1) Davis and Peet's (1976) "minimum area" spatial 
filter modified to accept apriori class conversion weights: 2) The 
'IBIS' procedure which is actually several variations involving image 
magnification, a modal filter, and individual class conversion weights; 
and 3) A voting rule technique for labeling existing manually delin 
eated timber stand maps (Woodcock, et al., 1980). 
Post-Processing Issues. There are two issues which need to be con 
sidered in selecting a post processing procedure. The first issue is 
whether diagonal pixel connections are to be allowed, and the second 
issue relates to a comprehension of the spatial displacement and 
class frequency histogram change that occur as a result of post 
processing. 
The diagonal issue is of concern when classified image data is to be 
converted into vector format. Diagonal connections permit polygons to 
effectively across over each other and occupy a cannon location 
(figure 1). This occurrence is an artifact of using conceptually 
square pixels in image processing, and results in cctnplex polygonal 
maps that can be misleading and difficult to use. The additional use 
of a "post-thinning" algorithm to smooth the pixel "stair-step" effect 
enhances the location of the diagonal cross-over points by preserving 
geonetric shapes at that point, and often results in an oddly appear 
ing map of mixed round and angular shapes. 
If the final map product is to be left in image format, then the diag 
onal issue will probably not be of great concern. Conversion to vector 
format for incorporation into polygonal geographic information systems, 
however, may result in difficulties with polygon overlay, area measure 
ment, labeling, and visual interpretation. Depending on the applica 
tion, the vector geographic information system, and the alternative, 
diagonal connections may be acceptable. For the FOCIS project, diag 
onal connections were permitted, and the modified Davis and Peet 
spatial filtering technique used. It is the purpose of this paper to 
report the methods and alternatives that were investigated in the 
course of selecting an appropriate post-processing technique.