Temporal Texture Data

I have collected a database of video clips of temporal textures, such as moving water, steam etc. Based on such real examples I have generated synthetic temporal textures.
They were generated using the spatio-temporal autoregressive model (STAR), which is described in the following articles:

Martin Szummer, Rosalind W. Picard. Temporal Texture Modeling September 1996 Lausanne, Switzerland IEEE Intl Conf On Image Processing (ICIP) pp. 823-826 vol. 3
Martin Szummer Temporal Texture Modeling (September 1995). This is a more in depth description and is a MIT Media laboratory tech report 346.

Sample images from the video sequences


River	Boiling water	Steam	Swirling water

Image Sequence Download Instructions

Download temporal textures here.

Unfortunately, the file describing the file format got lost, so there is no information about the format of the data. It is 10 years since I used them, and I only recall they in the Media lab “DAT” homegrown format which I cannot find any information about. I looked around among my old files, and I found some Matlab code to read and write DAT files. Unfortunately, it only seems to support 1 channel files (no color). The Matlab files are here – it is clear that the greyscale reader is just reading raw bits.

I am not sure about the format for the color. From the description file we know that they are dimension 256x256, 3 channels and 100 frames.
The color format: (_data_sets 100) (_channels 3) (_dimensions 256 256) (_data_type unsigned_1)
There are some obvious things to try. Either RGB RGB pixelwise, or RRRR GGGG BBBB imagewise, or perhaps RRRRR GGGG BBBBB for each dataset.

If you try it, and let me know when you figure it out so I can create a new README file?

Reference

I would be pleased if you could reference my publication in any work you publish which uses this data (as it took me time to collect the data).
Here is a reference in BibTex format:

@INPROCEEDINGS{Szummer96a,
  author = {Martin Szummer and Rosalind W. Picard},
  title = {Temporal Texture Modeling},
  booktitle = {{IEEE} Intl. Conf. Image Processing},
  year = {1996},
  month = sep,
  volume = {3},
  pages = {823--826},
  url = {http://research.microsoft.com/~szummer/papers/icip-96/SzummerPicard-icip96.pdf},
}