Continuous Generalization for fast and smooth Visualization 
on small Displays 
Monika Sester & Claus Brenner 
Institute for Cartography and Geoinformatics, University of Hanover, Appelstrafe 9a, 30167 Hanover, Germany 
monika.sester,claus.brenner@ikg.uni-hannover.de 
Commission IV, WG 1V/3 
KEY WORDS: Cartography, Generalization, Mobile, Real-time, Visualisation 
ABSTRACT: 
With the increasing availability of small mobile computers, there is also an increasing demand for visualizing cartographic objects 
on those devices. Prominent applications are location based services in general, and car and pedestrian navigation in particular. In 
order to be able to offer both detail and overview of a spatial situation, the devices have to provide flexible zooming in and out in 
real-time. The presentation of spatial data sets in different zoom levels or resolutions is usually achieved using generalization 
operations. When larger scale steps have to be overcome, the shape of individual objects typically changes dramatically, also objects 
may disappear or merge with others to form new objects. As theses steps typically are discrete in nature, this leads to “popping 
effects” when going from one level of detail to the other. 
In this paper, we will describe an approach to decompose generalization methods into elementary operations that can then be 
implemented in a continuous way. For example in the case of displacement, an object will not simply jump from one position to the 
other, but slowly shifted from its original position to the new one. In the case of simplification of building ground plans, the 
clementary operations e.g. care for removing extrusions or intrusions of buildings, as well as offsets. In the paper we will identify 
elementary generalization operations and also present their implementation as a continuous operations. We will apply these concepts 
for line generalization, the generalization of building ground plans and for displacement. 
1. INTRODUCTION AND OVERVIEW 
The presentation of spatial data in different levels of detail is a 
basic requirement in order to be able to fully understand spatial 
processes. [n cartography it has traditionally been accounted for 
by the series of topographic maps (e.g. different scales from 
1:10.000 to 1:1 Million). For their production, generalization 
operations are being applied that generate a new representation 
from the given detail data. 
The need for presenting spatial data in different resolutions 
recently came up again from a completely new domain: in order 
to present spatial information on small mobile displays — 
typically user location or navigation instructions — there is a 
strong need for generalization, because on the small displays 
only a reduced information content can be displayed at a time. 
As the small display devices typically do not dispose of large 
storage capabilities for storing digital data sets at different 
resolutions, the need for efficiently transmitting the spatial 
information arises. 
This was the basis for this research, that aims at developing a 
method for incrementally transmitting more and more 
information in terms of object details to a small mobile device 
through a possibly limited bandwidth channel by incremental 
streaming. When a user inspects spatial data using a mobile 
client, first only the coarsest information is transferred to give 
an overall impression. Then, objects in the vicinity of the user 
will be incrementally loaded, until — if the user wishes so — the 
whole scene is given at the highest level of detail available. 
The idea is to pre-compute a sequence of vector representations 
at different levels of detail, which are then incrementally sent to 
the client. These different representations, in our case, are 
coded efficiently in terms of a set of simple operations. These 
operations can be generated by appropriate adaptation of 
existing generalization operations. The procedure provides 
methods to visualize and animate these changes in a continuous 
and streaming fashion. 
The paper is organized as follows: After an analysis of demands 
for progressive information transmission, a brief classification 
of generalization algorithms is given. Then, the elementary 
operations to code incremental changes of objects are presented. 
Different generalization functions are adapted in order to 
produce a representation in terms of those simple operations. A 
summary concludes the paper. 
2. RELATED WORK AND DEMANDS FOR 
PROGRESSIVE INFORMATION TRANSMISSION 
The basic demand for continuous generalization is that the 
changes occurring when going from one representation to the 
next are small enough in order not to be noticed. Thus, the user 
is not disturbed by coarse visible changes like object parts 
popping up or objects suddenly disappearing. 
In order to provide such a smooth transition from one scale to 
the next, incrementally representations with more and more 
detail have to be visualized. This would imply that a very dense 
series of different representations is generated and has to be 
transmitted to the user while he/she is zooming in or out. 
1293