Vis5D Enhanced with Feature Tracking

Introduction

FVis5D is a modified version of Vis5D that provides interactive feature based visualization of large 3D time varying datasets. It integrates our feature extraction and feature tracking algorithms with Vis5D. It supports the visualization of any Vis5D file, and provides feature based visualization techniques while maintaining all original features of Vis5D, in the hope that it will provide scientists with more effective ways of understanding their large 3D time varying datasets. The feature tracking addition colors isosurface-features the same throughout their lifetime, i.e. a yellow feature in timestep 100 is the evolution of the yellow feature from timestep 1. (NOTE: color is now associated with different features of the same isosurfaces, not only different isosurfaces.) Quantifications are also computed and available for analysis, e.g. volume change of a feature over time, number of features over time, etc..

Below is a screen shot of the modified interface that enables the display of color encoded isosurface rendering with feature tracking information. The additional data column (highlighted with red arrow) is the track column which brings up its own threshold slider. It works like the isosurface column except that now features are tracked when animate is activated. In the figure, the isosurface of the CLW variable is shown and the different connected components (features) are colored differently (NOTE: the color in this figure does not refer to different thresholds or different variables. For more information about feature tracking please click here.

How to get it?

If you haven't installed Vis5D, you need to install it first. Download the official Vis5D (version 4.3) distribution from Vis5D home page, which contains LUI user interface library needed for the compilation. Put it in the directory where you install Vis5D.

Precompiled FVis5D executable is currently available for SGI only. You can just download it into directory where you install Vis5D, and uncompress it with the command:
%gunzip fvis5d.gz

We still need access to other platforms to expand the list of binaries. However building the executable should be pretty straightforward.

How to install it?

Currently FVis5D only works on the systems with OpenGL or Mesa. We will try to port to other platforms, if we gain access to them.

Here are the installation instructions:

1. Go to the directory where you have the Vis5D source.
For example,
% cd vis5d-4.3

2. Uncompress and un-tar the archive file:
%gunzip -dc fvis5d-1.3.tar.gz | tar -xvf -

3. Compile your vis5d program if you haven't. See the Vis5D installation guide.

4. Compile FVis5D
%cd Fvis5d/Track
%make
%cd ..

Run make with appropriate make command as you compile Vis5D. For example, suppose you have an SGI Iris 6 with OpenGL and 3-D graphics hardware. You should type:
%make irix6

How to run it?

Before you start

If you are not familiar with the concepts of the feature based approach for the visualization of time varying datasets, we suggest you browse our feature tracking web site and go through related papers before you use the software.

A feature based approach performs segmentation to extract regions above a selected threshold, and then uses feature tracking techniques to compute regions and their properties, and follow their evolution over time. The tracking information provides feature evolution information (i.e. feature A from timestep 1 is the same as feature B from timestep 2). Isosurfaces are assigned colors to match properties of the object evolution. In this implementation, a particular region in all the time steps is assigned the same color. When features merge, the feature with larger volume establishes the color.

Starting fvis5d

Using fvis5d is pretty much the same as vis5d. After you have made a v5d file, run with the command:
fvis5d file.v5d [options]

All options of Vis5D are still applicable. Two options are added for feature tracking:
-ctable color_table_file
Use the customized colors defined in the color table to color the new features.

-verbose
print tracking information.

The Control Panel

The last column of buttons in the control panel's button matrix (annotated by a green dot) controls the extraction of thresholded regions. In current implementation, voxels above the selected threshold are extracted, and isosurface is then fitted around the thresholded region for visualization. (We will try to implement more general definition of the thresholded regions with interval volumes or regions below threshold in the future release.) Therefore currently volume inside the isosurface contains voxels with values above the threshold. The segmentation routine also has the option to filter small objects which are usually "noise" or unimportant features. Feature tracking algorithm is used to automatically correlate these regions over time. When a 3D variable is selected with the left mouse button, a pop-up window with a slider and OK button appears below. Select a threshold on the slider, type in the minimum volume for filtering small objects ( Default is zero, no filtering), and then click on the OK button to trigger the object segmentation and feature tracking routine to generate isosurface and color them with feature tracking information. Click on a track button with the middle mouse button will activate the pop-up window without toggling the variable on or off. NOTE: To assure correct feature tracking result, you have to make sure that the time interval is small enough to guarantee the overlapping of regions between time steps. You can verify this with visual testing on a few time steps. While regions are not correlated correctly, you have to use smaller time intervals.

Toggling ANIMATE on will let you observe the evolution of features with color-encoded history information. Note that the feature tracking routine is running asynchronously, so you may not see the updated results while it is in progress. Once it is finished, you will see an output "tracking done" in your console window where you start the fvis5d.

Isolating Features

Click on ISOLATE button in the control panel denoted by a red dot will open a window in which you can specify which object starting at which time step you want to isolate. ( Ideally, one should be able to interactively pick a object to identify its Obj_ID in the 3D image window. Unfortunately the user currently have to guess this number, but you may get some clue by looking up the id by its properties specified in .attr file. We are currently implementing pick functionality which will greatly ease this task. ) Feature isolation can be performed either forward or backward. If the evolution of only one feature is of interest, it can be isolated ( forward-isolation ) and highlighted. The other features are removed. When a feature's history is desired, backward-isolation can be used. In forward isolation, the first dataset contains only one feature and subsequent datasets may contain more. In backward isolation, the last dataset contains only one feature and the first may contain many.

Customized Color Table

Bounded surfaces are colored coded with feature tracking information. By default, a newly created object is assigned a random color. Users are also allowed to specify the color these newly created features with their customized colors, which are defined in the color table. If the number of newly created features exceeds the number colors defined in the color table, random color will then be used.

The first line of the color table is number of entries. The rest lines are values of red, green and blue components in the range of [0, 255], separated by a space, and each line is terminated by carriage return.

Format of Color Table File:
   Number of colors (1 integer)
   red green blue (3 integers, separated by space)
   red green blue (3 integers)
   ....

An example of the color table is,
% more color.table
6
255 100 0
100 100 100
0    255 255
255 255 0
255 255 255
255 0   0

Output of the Program

Object Attributes:
FVis5D computes the quantification of features for all time steps. All the attributes are stored in the attributes file (.attr). The name of the attributes is formed with the based name of v5d file and time steps. For example, your v5d file is mm5.v5d. the set of the attribute file of mm5_1.attr, mm5_2.attr ...

will generated in your working directory.

An example of the attribute is as following:
Time step: 25
Date: 34040
Time: 00:37:80
Threshold: 0.000296
Total objects found: 5

________________________________________________________________________________

Object 1 attributes:
        Max position: (30,11,2) with value: 0.000495
        Node #: 26
        Mass:   0.009868
        Centroid: (28.983986, 10.812018, 2.343744)
        Moment: Ixx = 0.904073
                Iyy = 3.110146
                Izz = 0.225584
                Ixy = -0.734182
                Iyz = -0.479759
                Izx = 0.140511
        Local Max:
            max-1: 0.000495 position (30, 11, 2)
________________________________________________________________________________

...
...

FVis5D can also generate quantification plots for time-varying datasets. Currently, the plot of number of objects vs time is available. The default file name is xxxx.num, where xxxx is the basename of .v5d file. The .num file consists of two columns. The first column is the timesteps, the second column is the number of objects in this specific time step. You can essentially view this plot with any plot program. For example, xgraph with the following command:

%xgraph -bb -tk -x "time" -y "Number of Objects" xxxx.num

Disclaimer

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANT ABILITY or FITNESS FOR AND PARTICULAR PURPOSE. As such we do not provide real support. However, we will be happy to answer questions and help with minor problems (preferably by email).

Future work:

This is our first release of FVis5D (Version 1.0). We are still working on incorporating many of the enhanced interactive visualization techniques into FVis5D.

Features to be included:

Pick functionality, which allows user to pick a feature of interest.
To display plots of quantification of the feature of interest.
To render the isolated regions with volume rendering incorporated with feature tracking information.

Vizlab

You are here