CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
of compressed error-image and the bytes needed to code 
the projective transformation parameters. 
Table 1 : Results of the frame compression algorithm 
Polygon 
Compr.sed 
Error-image 
(bytes) 
Compr.sed 
View B 
(bytes) 
Bit/ 
pixel 
error 
Bit/ 
pixel 
view 
B 
Ratio 
Green 
625 
780 
0.07 
0.08 
1.25 
Red 
603 
768 
0.06 
0.08 
1.27 
Yellow 
586 
753 
0.058 
0.079 
1.28 
It should be noted that the compression ratio is always greater 
than 1, what confirms the effectiveness of implemented 
compression procedure. Indeed, this means that sending to 
the client both the compressed error-image and the 
parameters of the projective transformation, generates a 
throughput lower than the transmission of the compressed 
view only. As shown in table 1, the gain is ranging between 
25% and 28% in terms of bytes employed. 
6. CONCLUSIONS 
In this paper an innovative concept of VRML browser has 
been proposed. It is aimed mainly to allow the display of 
complex and large 3D models, as the ones obtained by 
ground-based laser scanners, reducing the computational load 
on the user-side. To this end, we developed a VRML split- 
browser, which is conceptually based on a server/client 
paradigm, but at the same time it keeps all the advantages of 
the VRML interchange file format as suitable mean to share 
3D objects along the web. The main idea on the ground of 
developed architecture is the splitting of the tasks between 
server and client. The former performs all the operations 
requiring the most computational effort, reducing the 
working load of the latter. This has been accomplished using 
perspective transformations and image compression 
algorithms, like LZ77. In our system the server sends to the 
client only a set o parameters, defining the perspective 
transformation between to 3D scenes, and a compressed 
error-image, instead of the whole VRML 3D model. Then the 
client uses such elements to reconstruct the view of the scene 
according to the input command activated by the user on the 
client’s GUI. 
The proposed work is still in progress, as improvements can 
be still applied, in order to further optimize the performance 
of our split-browser. We are investigating the use of the 
Bounding Box, view-images caching and the adoption of a 
linear prediction scheme in the LZ77 algorithm. The former 
would allow to compress only the portion of the whole 3D 
model that is actually enclosed by the bounding box, saving 
in this way the number of bytes used for the image coding. 
The second solution deals with the creation of a memory 
buffer on the client side, where all the views sent by the 
srever are sequentially stored. In this way, if the eye-point 
(user’s viewing direction) returns back on a previous visited 
position, the client needs to restore the corresponding view 
from the buffer: no data have to be sent from server to the 
client. A further server/client throughput reduction could be 
achieved by means of the linear prediction: the pixel color of 
an image can be predicted on the baiss of the color of the 
neighbours. Accordingly, the server would not compress the 
original image but rather the so called residual-image, whose 
pixels are obtained as difference of the color values of 
adjacent pixels ont he source image. 
ACKNOWLEDGEMENTS 
This work was developed with the project “Application in the 
survey, store and management of environmental and cultural 
resources of GNSS/INS positioning and satellite, aerial, 
terrestrial photographic and laser scanning data, transmitted 
by DARC, GSM/GLOBAL STAR, INTERNET methods”, 
partly financed by MURST (Italian Ministry of University 
and Research) in 2002 as project of relevant National interest. 
National coordinator: Giorgio Manzoni, head of the Research 
unit Antonio Vettore. 
REFERENCES 
Carey R., Bell G, 1997. The Annotated VRML 97 Reference 
Foley J., Dam A. V., Feiner S., Hughes J., 1996. Computer 
Graphics: Principles and Practice. Addison-Wesley, 2 nd 
edition. 
Gleicher M., Projective Registration with Difference 
Decomposition. URL: http/Avww. cs.wisc.edu/Graphics/ 
Papers/Gleicher/Califomia/track-Final.pdf 
Hartley R., Zissermann A., 1999. Multiple View Geometry. 
CVPR, June. 
Mason W, 1997. Open-GL Programming Guide. Addison- 
Wesley, 2 nd edition. 
Mian G.A., Rinaldo R., 2000. Elaborazione e Trasmissione 
delle Immagini. Ed. Libreria Progetto, Padova. 
Nelson M., 1992. The Data Compression Book. M&T Books. 
Ziv J., Lempel A., 1977. A Universal Algorithm for 
Sequential Data Compression. IEEE Transactions on 
Information Theory, n. 23, pp. 337-343.