DESIGN AND IMPLEMENTATION OF A GIS BASED DIGITAL IMAGE PROCESSING SYSTEM 
M. Halim 
Universal Systems Ltd. 
Fredericton, N.B., Canada 
E. Derenyi, R. Pollock!, T.Y. Shih2, C.K. Xian 
Department of Surveying Engineering 
University of New Brunswick 
Fredericton, N.B. Canada 
ISPRS COMMISSION IV 
ABSTRACT 
An increasing demand for digital images as an input to geographic information systems (GIS) creates a need for a data processing 
environment which incorporates the main functionality of stand-allown image processing systems while maintaining the full 
capabilities of a GIS. This paper presents the design and implementation of the Computer Aided Resource Information System 
Raster Image Extension (CARIS/RIX) which is such an environment. The main purpose of CARIS/RIX is to facilitate the 
extraction of information from digital images while displayed in registration with digital maps. Certain image processing and 
analysis functions have been developed which are specific to this new environment and a number of conventional image processing 
operations have been included. A user server interface facilitates future extensions of this facility. 
KEY WORDS: Hybrid GIS, Data integration, Image processing, Image display, Integrated System, CARIS. 
INow with the Department of Computer Vision, University of British Columbia, Vancouver, B.C., Canada. 
2Now with the Department of Civil Engineering, National Chiao-Tung University, Taiwan, R.O.C. 
1. INTRODUCTION 
There is a rapidly increasing demand among users of 
geographic information systems (GISs) for the use of digital 
images as a source of data. This demand can best be served 
by a fully integrated GIS and image processing system 
(IPS). Accordingly, in the early part of 1989, researchers at 
the Department of Surveying Engineering, University of 
New Brunswick (UNB) began to develop a series of image 
analysis software modules that could access the spatial and 
attribute information stored in a GIS. The project resulted in 
a series of software packages which support the commonly 
used image processing and analysis functions and operate on 
a raster data structure which is compatible with the existing 
data structure of the GIS. The result is CARIS/RIX the 
Computer Aided Resource Information System (CARIS) 
Raster Image Extension (RIX). 
2. CARIS SOFTWARE OVERVIEW 
CARIS is a comprehensive, vector based, GIS developed 
and marketed by Universal Systems Ltd. (USL) in 
Fredericton, N.B., Canada [Masry, 1982]. It runs under 
both the UNIX and VMS operating systems and supports a 
variety of input and output devices. The user interface is 
based on X-windows and MOTIF standards. It consists of 
three comprehensive software packages as illustrated in 
Figure 1 [USL, 1991]. Each package contains a set of 
software modules which perform specific functions. 
The Cartographic Processing Package is the entry 
level package. It supports the collection and editing of 
cartographic data and the creation of high quality 
cartographic output products. Central to this package is the 
Primary Graphics software module, which comprises an 
interactive graphics editor, CARED, and supports 
commands to capture, modify, maintain, and display 
cartographic data. 
The Topology Creation Package provides the creation 
and maintenance of topology and attribute information, 
which may be attached to the graphical data. It also supports 
the formation and maintenance of polygon and network 
topology. 
The Database Analysis Package is the top level 
package, which serves proximity analyses, database 
inquiries, and report generation from a CARIS database 
776 
consisting of graphic and attribute files. CARMAN, the 
CARis MANager, is used for querying the graphic database 
and spawning queries to the attribute database. Links 
between graphics and attributes are established by an 
alphanumeric key. 
In addition, the Specialized Support Application Package 
serves a number of optional operations. A development 
library is also provided. 
3. IMAGE DATA ORGANIZATION 
3.1 The Raster Data Structure 
During run-time, the pertinent raster data is stored in a 
system of tilemaps [Reedijk, 1990]. A tilemap is an array 
based representation of a 512 by 512 pixel portion of the 
raster extent. By dividing the raster into several manageable 
pieces, an efficient processing mechanism is achieved 
without demanding a great deal of storage space. In fact, 
this scheme allows the interactive processing of the raster 
data in random access memory (RAM). For a permanent 
representation of raster data, a storage format is used. 
The maintenance and handling of the tilemaps is the 
responsibility of the tilemap manager. Many tilemaps may 
reside in memory during run-time and the tilemap manager 
swaps the tilemaps as required. Associated with each 
tilemap is its geographic position within the raster extent, the 
size of the pixels in ground units, and the number of bits 
assigned per pixel. These values constitute the so-called 
raster definition file. 
A raster image is viewed as consisting of one or more layers 
with each layer consisting of one or more regions as shown 
in Figure 2 [Derenyi and Pollock, 1990a; 1990b]. In a 
region, all pixels must be rectangular and must correspond to 
the same raster definition. This means that they are arranged 
according to a uniform rectangular grid, they have the same 
ground unit dimensions, and their values are stored using a 
uniform number of bits. The extent of the region and the 
ground location of each of its pixels is defined to an 
acceptable accuracy, or can be derived from information 
stored in association with the region. 
An example of a region is a single channel of a full Landsat 
Thematic Mapper (TM) scene. Each pixel covers a nominal 
30 metre by 30 metre area on the ground and is stored in 8 
  
  
  
FA. ps US n ain C CU) ia a 
Fri uu c ANA - 0 M omms rm om