24 
3.2 OPERATIONS WITHIN THE VARIOUS VIEWS 3.3 
This section gives a brief overview of the types of operations that are defined within each view. Specifically, the Some 
retrieval operation is considered, and the various retrieval qualifiers that may be applied within each view are given. consi 
For simplicity, temporal and error characteristics of the data are for the present ignored. foresi 
foresi 
field, 
3.2.1 Operations on the Field and Image View ar bi tr 
The field and image views contain spatial information in the form of graphical primitives (points and pixels co-in 
respectively) and the property values occurring therein. The data has no higher level identity imposed upon it, so 
operations are limited by the lack of feature identifiers or class labels on any of the graphical primitives. A retrieval 0ne ( 
qualifier applied to field data must therefore either address all primitives without restriction, for example window(), t£ 
rotate(), or must first select a subset of primitives based on their available properties. These properties are of two must 
types: geometric (positional) and value (attribute). datas 
It is also theoretically possible for other geometric types to exist at this level, that is without an associated 
interpretation or meaning, but it is not possible for an object at this level to possess any properties used to denote 
features or classes. 
3.2.2 Operations on the Thematic View 
The thematic view manipulates classes. It regards spatial data in the form of graphical primitives where the value 
properties of each primitive have been replaced with a single class label. Retrieval qualifiers may still address the 
underlying geometry, but now instead of the original data values, only the properties of their labelled class may be 
used. Note that there is now no flexibility as to how the graphical primitives can be selected by value, since an 
interpretation scheme has already been imposed. New operations become possible by way of the thematic nature of 
the data; for example, selecting a class, merging two classes and calculating the area occupied by a class. 
3.2.3 Operations on the Feature View 
The feature view manipulates explicitly defined geographical features. It provides a view where the data values or 
class labels have been replaced with feature identifiers. As before, operations can address the underlying geometry 
or any other property of the feature, via the feature identifier. The feature identifier also provides access to any 
additional descriptive data which may also be used to impose retrieval qualification. Note that there is no flexibility 
as to the type or the shape (spatial description) of the feature, since both are now defined explicitly. 
Table 1 shows how the three views of data outlined above affect the basic unit of manipulation in an operation. 
Nowhere in the above descriptions is any mention made of data type (raster, vector) or data organisation. All views 
may use as their basis any of the supported spatial data structures. In practice, the preferred types may well be raster 
for image data in the geometric view, quadtree for the thematic view, and topological vector for the feature view, 
but the user need not be concerned with this as it does not impinge on their interaction with the system. 
Table 1: Summary of abstraction and operation across views 
Unit of Manipulation 
Valid Retrieval 
Qualifiers 
Field View 
individual data objects 
point/pixel geometry 
point/pixel values 
Thematic 
sets of data objects 
class geometry 
View 
describing a class 
class attributes 
Feature 
sets of data objects 
feature geometry 
View 
describing a feature 
feature attributes 
The < 
these 
may 
struci 
aggre 
estim 
deriv 
their 
imag 
In fo 
data 
effec 
assoc 
serie: 
repre