Advanced Multimedia Processing Lab -- Projects -- Progressive 3D Model

            About AMP Lab        Projects        Downloads        Publications        People        Links

Project - Progressive 3D Model

Contents

 

Team Member

Masahiro Okuda

okuda-m@kitakyu-u.ac.jp

Top of this page

                                  

Goal

The 3D models usually have one or more attributed data such as normals, colors, or textures.  If one really wants the 3D models to look authentic, texture maps are required. In this project, we seek to address the flaws of conventional 3D format such as VRML and propose a joint geometry/texture codec for the 3D models rendered with textures, resulting in  progressive bitstreams.  In our codec, the most important bits to represent the 3D models are sent before the less important bits, allowing the users to stop the download at any time they want. We are also engaged in view-dependant progressive coding, animation coding and error resilient coding of 3D models.

Top of this page

                                         

Motivation

With the rapid growth of computer and information technology, 3D modeling capabilities are becoming an increasingly important part of many fields, including CAD/CAM, medical imaging, game design, and so on.  Computer graphics often requires complex, highly detailed models to maintain a convincing level of realism with VRML having become the de facto standard for high-quality 3D graphics representation. With the proliferation of many excellent VRML viewers (such as CosmoGL by SGI), there has also been a surge in VRML content that is freely available. However some major drawbacks to the VRML standard have begun to emerge, such as the fact that VRML files are bulky and time-consuming to download. This is because, in an effort to keep the format compatible across the various computer platforms, they have kept to the verbose ASCII format. In addition, VRML browsers require the complete file to display the models even when viewers are interested only in a part of or a low-resolution version of the models. Since the computers used to view the file range from low-end PCs to top-of-the-line supercomputers, and since the files are sent over slow telephone line as well as faster ISDN, the level of detail actually necessary may vary considerably. Thus the progressive coding and transmission satisfy the variety of demands.

Top of this page

 

System Description

The algorithm is based on vertex decimation and texture re-mapping. The encoder combines a progressive geometry scheme and progressive texturing and send more important bits before less important bits.

We created a  browser which decodes the progressive bitstreams and displays the models progressively. Here is what a model might look like as it is being loaded. The first, fifth, and last layers were obtained by decoding 15%, 22% and 100% of the total bitstream respectively.

First layer

Fifth layer Last layer       

                            

Demo

- Download Progressive 3D browser now!

- For setup instruction, please see FAQ.

- Example files 

Try out progressive streaming (short text files that contain a URL):

[3D Model without Texture]

Filename Description
beethoven.gtl Bust of Beethoven
opt-eight.gtl Figure of Eight
opt-femur.gtl Femur
opt-horse.gtl Horse (no tail)
opt-pieta.gtl Famous statue
opt-shape.gtl Star-shaped object
opt-skull.gtl Skull
triceratops.gtl Triceratops
cow.gtl Cow
croco.gtl Crocodile

  [3D Model with Texture]

     Filename Description
duck.gtl Duck 
vase1.gtl Vase 1
vase2.gtl Vase 2
totem.gtl Totem Pole

 

The following files are the actual files, if you want to view them offline

[ 3D Model without Texture]

Filename Description
beethoven.jgt Bust of Beethoven (75KB)
opt-eight.jgt Figure of Eight (22KB)
opt-femur.jgt Femur (108KB)
opt-horse.jgt Horse (no tail) (309KB)
opt-pieta.jgt Famous statue (97KB)
opt-shape.jgt Star-shaped object (71KB)
opt-skull.jgt Skull (306KB)
triceratops.jgt Triceratops (79KB)
cow.jgt Cow (86KB)
croco.jgt Crocodile (300KB)

 

  [ 3D Model with Texture]

     Filename Description  Original File*
duck.jgt Duck (453kB) 605kB
vase1.jgt Vase 1 (433kB) 651kB
vase2.jgt Vase 2 (462kB) 651kB
totem.jgt Totem Pole (465kB) 683kB

                   * The original VRML files and texture files are compressed by gzip and jpeg, respectively

The original VRML files with textures can be found in http://www.vit.iit.nrc.ca/3D/Pages_HTML/3D_Models.htmll

Top of this page

 

Our work is used by...

Top of this page

 

Publications

Top of this page

 

Contact

Any suggestions or comments are welcome. Please send them to Masahiro Okuda 

Top of this page