Issue: Volume: 24 Issue: 4 (April 2001)

Boujou 1.0

By George Maestri

Any digital artist who works with live-action footage has probably had to perform the nasty task of matchmoving. In theory, matchmoving is simple-in order to produce a realistic composite of 3D imagery and live-action footage, you match your 3D software's virtual camera to the real camera used in the shot. In practice, however, matchmoving can be a nightmare, because first you need to deduce where the real camera was located and what type of lens was used when the footage was shot. After that comes the arduous task of matching the virtual camera frame by frame to the real camera.

Over the years, software manufacturers have developed a number of solutions for the problem of matchmoving, and most of these have gone a long way toward restoring artists' sanity. 2d3's boujou is the latest such solution.

Software installation for boujou is straightforward, as is the program's interface. Along the top is a menu bar with icons for major functions such as Load Sequence and Play. Boujou provides a large window for viewing footage, and you can split the window to view multiple frames simultaneously. The Tasks window provides information about the status of the tracking process.
Boujou's interface is straightforward, making this matchmoving application easy to use.

What makes boujou especially easy to use, however, is that it's almost totally automated. To matchmove a scene, you follow several simple steps.

The first is to import the film sequence into boujou. The software supports a variety of formats, including AVI, Targa, Cineon, JPG, TIF, and many more. (Conspicuously absent in the review copy, however, was support for QuickTime movies. According to 2d3, this feature will be added in Version 1.1.) The software provides a dialog box for user-supplied information about the camera, such as lens focal length. Although optional, such information will be available to users in most production situations, and providing it to boujou helps speed up the process.

After the sequence is loaded, it is displayed on the screen. Before boujou can begin tracking, you need to tell it what to track. The software guides you through this process by flagging steps with diamond-shaped icons. Three steps-Track Features, Track Camera, and Export-are flagged with three diamonds because they are critical to the process. Other steps, such as Describe Cameras, Import Image Hints, and Visualize Tracking Results, are marked with one or two diamonds and are optional.

The next important step is tracking the features of the scene, which boujou does automatically. First, it selects a large number of points to track. As the software works its way through the scene, it begins to discard points that produce anomalous results, retaining only those that produce a consistent track. For instance, the test scene used for this review was a handheld shot of a swimming pool. The flickering reflections on the water were initially tracked but soon discarded in favor of the edges of the pool, which provided the best tracking data.

After the features are tracked, this information is used for the next step, which is to extract the camera information. This includes the focal length of the camera and its position within the scene. As noted earlier, if you know any of this information, you should give it to boujou when the sequence is first loaded, as this will speed the process and improve accuracy. Still, in the test scene, boujou proved accurate with no user input whatsoever.

After the camera is tracked, the final step is to export the camera tracking information for use in Discreet's 3D Studio Max, Alias|Wavefront's Maya, Softimage, or ASCII formats. I imported the information into 3D Studio Max, and the process was straightforward; boujou simply generated a Maxscript file. Loading and running this script created the tracked camera and the markers in the scene.

To test the track, you just load the image sequence as a background plate. A 3D object placed in the scene should appear stable against the plate when viewed through the tracked camera. In the case of the test scene, it was easy to model and place a beach ball in the pool. However, I found that the scale was a little off, in that it seemed as though the scene generated by boujou was physically smaller than the one I intended to match. Fixing this required scaling the object to fit.

Overall, boujou is the easiest matchmoving tool I've ever used. Most packages, such as RealViz's Matchmove, require users to manually define stable points in a shot to help with the process. By performing this task automatically, boujou saves a lot of time and headaches. This is an excellent tool for anyone who needs to match live action and CG.

George Maestri is a writer and animator living in Los Angeles.

Price: $10,000
Minimum System Requirements: Window NT 2000; 256MB of RAM