RFAI database of
3D synthetic textured images



Why creating this database of 3D images ?

The RFAI 3D image database is a collection of synthetic textured 3D images.
It is maintained primarily to support research in 3D image analysis, 3D image classification, and machine vision.
For the moment, each image is made using only one texture model so only classification tests are possible. The available images are not allowing segmentation contests.

This is the first edition of the RFAI 3D image database and many new images should be added in the futur (for segmentation purpose for exemple).

This first edition has been done in the context of Ludovic Paulhac's PhD thesis. The objective of this work is to propose and evaluate different methods of 3D images characterisation using original 3D texture features.

We have decided to provide free access to this database in order to allow benchmarking between the diferent proposed methods of 3D texture characterization.
We haven't find any other such databases of 3D images so we hope this one will be usefull for researchers interesting in 3D images characterisation.

Commercial distribution or any act related to commercial use of this database is strictly prohibited.


Description of the RFAI database :

As said before, for the moment, the available images are not allowing segmentation contests because each image is made using only one texture model. So only tests of characterisation and classification are possible.

The images have been constructed using two different methods:
  • With the first method, the three-dimensional textured images were constructed using two-dimensional texture images like Brodatz textures, fractal textures etc. Two or more two-dimensional texture images are interpolated to obtain a three dimensional image.
  • The second method consists to use geometric shapes like sphere, cube, ellipsis, etc. For exemple, to construct a gruyere texture, we place randomly white sphere with random sizes in a black three-dimensional image.
  • It is also possible to mix these two methods that is to say to place textured shapes randomly in a textured image of the first class. We then obtain a third category of three-dimensional images.
  • To complete this database, we apply some transformations on each class of texture with translations and rotations according x, y and z axis.
  • Moreover, Gaussian noise and Gaussian blur are applied to increase variability in each existing three-dimensional texture class.
In order to easilly evaluate the diffierent characteristics of a characterisation method, the database is divided into sets of images based on the basic character of the pictures.
Each of these dasasets contains 30 classes with around 8 examples for each of them.
Images in each set are of the same size such as 64x64x64 voxels or 128x128x128 voxels. All images are 8 bits/voxel stored using a sequence of jpeg images.

The following sets are currently available (each one containing degraded images):
  • 3D textures from interpolation of 2D images
  • 3D textures from random insertion of 3D shapes
  • 3D textures from radom insertion of 3D textured shapes in textured backgrounds

Brodatz

Blobs with luminence distorsion

Brodatz with noise distorsion

Brodatz with random rotation


Downloading and condition of uses

Limitation of Liability

THE IMAGES CONTAINED IN THIS DATABASE IS PROVIDED 'AS IS' WITHOUT WARRANTY OF ANY KIND. THE ENTIRE RISK IS ASSUMED BY THE USER, AND IN NO EVENT WILL RFAI BE LIABLE FOR ANY CONSEQUENTIAL, INCIDENTAL OR DIRECT DAMAGES SUFFERED IN THE COURSE OF USING THE DATABASE.

Downloading and using the database
This database is publicly available and free for research staff.
All publications and works that use this RFAI database must reference the following paper: "Ludovic Paulhac, Pascal Makris and Jean-Yves Ramel : A Solid Texture Database for Segmentation and Classification Experiments, In 4th International Conference on Computer Vision Theory and Applications, pp 135-141. Lisboa, Portugal. February 2009.". [Download the paper]

You can also credit the "RFAI 3D textured images database project" as the source of the data by referencing this web page in your publication.

Also, please send a copy of your publication to JY Ramel. We will eventually put a list of references on this web page.

Permission to use but not reproduce or distribute the database is granted to all researchers given that the following steps are properly followed:
  • Send an e-mail to JY Ramel before downloading the database. You will need password to access the files of these databases.
  • All submitted papers that uses or talks about this database must cite the above technical report
  • Permission is NOT granted to reproduce the database or posted into any other webpage.
  • None economical profit can be obtained from this database.


Contacting RFAI

Technical questions about the database, the image format, or problems obtaining the images should be directed to the database editors:
  • J-Y Ramel, "ramel@univ-tours.fr"
  • Ludovic Paulhac, "ludovic.paulhac@labri.fr" (Deprecated)
  • Pascal Makris, "makris@univ-tours.fr"