Art of Tracking Part 2: Tips & Apps Overview

In part 2 of the Art of Tracking series, we take a look at the various software applications for 2D and 3D tracking. The software developers of these apps give us some technical insight as well as tips and suggestions how to get great tracks.

small
Click for Larger Image

2D Tracking

In looking at what tracks and how it is important to understand one key point, most modern trackers with the exception of Digital Fusion (and to a lesser degree Shake and AE) have all moved away from colour processing approaches. Almost all trackers are now biased to a luminance of black and white version of the track box. Digital Fusion alone amongst the leading packages uses colour information automatically. When a track is performed in Digital Fusion, the user interface shows the stength of each of the three colour channels and then which it has decided to use.

Digital Fusion also has another very innovative feature. While most 2D trackers also do corner pinning and 4 point tracking, DF offers a weighted solution. If you need to track a large object but not just one individual point on it and the object motion causes problems (such as tracking any object floating on water), DF allows several points to contribute to one solution. So even though an an individual point may have tracking errors due to perspective change, the overall movement of the object can be tracked from mulitple points and combined to provide one single track solution.

Shake also offers the ability to use hue or saturation for tracking instead of luminance. Ron Brinkman, author of ‘The Art and Science of Digital Compositing’, points out that in some cases of multicoloured patterns this can be very effective. While most tracks are correctly done using the normal tracker, if a monochrome version of your image has poor contrast, try a hue track. Adobe’s Michael Coleman notes that “After Effects uses luminance as the preferred color channel for the track, but there is a user control to switch to RGB or saturation. In practice, most hand-made tracking markers work best using luminance”.

Tracking now exists in every aspect of visual effects and is moving into colour grading. Discreet’s Lustre has tracking and Quantel’s QColor integrates into eQ and iQ to give multi-layer non-linear in-context color correction. Both applications bring effects tools to the colorist, which is something traditional color correctors haven’t offered.

What is the ideal tracking marker to use on set?

small
Click for Larger Image

We asked this question of almost every software supplier and technical expert. Their replies and tips may surprise you. “Well, for a start make sure your tracking markers don’t blow away”, jokes Quantel’s Steve Owen. “Framestore (has) a lovely story about a money shot in Walking with Dinosaurs. It was a helicopter fly past and they would have to comp in the 3d dinosaurs walking off towards their eventual extinction. They used baseballs as markers but unfortunately overlooked the down draught from the helicopter which promptly moved all the markers around. It was fixed with great skill in post.”

The common trackers are crosses, crash test dummy crosses in a circle, triangles or circles. The jury is out between the various experts, perhaps due to each product having a different algorithm. Coleman mentions that “AE doesn’t really prefer one object shape over another. I’ve always been pleased with results of using the cross or the dot, but this is probably personal more than technical. From a technical point of view, the AE matching algorithm has a technique for discovering the shape of the target. I believe that this technique is based off of the center of the feature region. This means that there is a benefit to putting the center of the feature region directly over the middle of the mark.”

The issue of which tracker is best is also effected by what the camera is doing – large dots zooming away to small pin points can be a problem, camera rotation, motion blur and depth of field can also be big factors. “Strictly speaking, a simple circle provides more robust tracking since it is rotation invariant � the tracking point remains the same even if there is rotation in the shot,” says Discreet’s Marcus Schioler. “Usually, losing a track occurs when the tracked shape changes or the lighting conditions change. Eliminating rotation variance is, therefore, a big plus. A circle also tracks better with zoom and focus changes. Curious Software’s David Franklin agrees. “The most important thing is that the shape of the tracking marker doesn’t change significantly during the motion. You want to minimise the alterations caused by rotation, perspective changes etc.,” says Franklin. “Therefore, I would say that a simple sphere or circle is best.” Digital Fusion’s Rony Soussan comments “let’s just say that there is no right or wrong answer, it all really depends on many factors. How close or far are the trackers, is the camera doing a dolly or pan? ..I’ve done music video’s where we had nothing but gaffer tape and we tracked it perfectly. It’s less about the tape or circle or square than it is about the artist driving the tracking software. Nothing’s automatic regardless off what people say. That being said, I’ve had a lot of success with a triangle myself, there a pattern that’s stronger then a circle so the tracker does have angles to track for scale as the marker changes size.”

A few years ago LED trackers appeared to be in vogue, but most of our experts agree LEDs do not make for better trackers, except in low light conditions “Providing the marker that you’re using makes a good contrast with the image, I wouldn’t expect LED to be better,” states Franklin. “Obviously the LED should have very good contrast, but I would be worried that glare/blooming from the LED would end up on the plates, making it harder to get a good result.” DF’s Soussan agrees and notes that he’s ” found that glowing or illuminated objects can give undesired results to the camera lens, like glowing and focus issues” However, Discreet’s Schioler points out that “using an LED marker could be a big help in some tough lighting conditions as the marker would remain visible even under dark shadows.”

While it is ideal to have contrast due to the fact that most trackers are monochrome, most experts agree there is little point in running special colour corrections. The tracking algorithms generally already have some sort of color correction or contrast stretching built-in. Ron Brinkman points out “there is nothing much you can do that the internal algorithms of Shake’s tracker can’t also do”. In fact, color correction done by an artist could have negative impact on the track by increasing the signal-to-noise ratio and interfering with tracking, says Discreets’s Schioler.

Dealing with Noise in Tracks

” (Noise….) there lies the art of tracking, ” jokes Digital Fusion’s Schioler. Fighting noise by reducing jitter or providing a cleaner track is a key aspect of doing successful tracks, especially in 4 point camera tracks. Tracks in their basic form can simply be edited by hand to remove spikes or errors. “Our tracker produces standard motion paths which can be used and edited anywhere in gFx” says Franklin. “You can remove jitter by deleting keyframes and allowing the curve interpolator to fill in the gaps, but we don’t currently have any specific tools in the tracker for removing jitter. We are adding these though.” Discreet already has a jitter removal option, introduced with version 4 of the software.

Most of the software experts we spoke with suggest experimenting with some presoftening of the clip prior to tracking. A small blur could reduce noise while maintaining the important image information for tracking. “AE can reduce noise and jitter in the track data using our “smoother” palette, which does just what it says,” according to Coleman. The AE expression engine can also be used to filter the track data. Curious Software’s Franklin has “found presoftening will often help with motion blur – the tracker will tend to follow the ‘center of mass’ if there’s some amount of blurring.”

Soussan aids the following suggestions for reducing noise, “One of the ways you can do it in Fusion, is simple to reduce points. Once you’ve tracked in Fusion it basically becomes a path that the tracker reads. All paths, and shapes in fusion, including effects mask, curves in your spline editor and the track data can be smoothed down using our reduce points system. It will average the shape of the track down as far as you want to push it until it’s a smooth spline curve. Another way is to de-grain an image prior to tracking, then replace back the original footage, I’m sure you’ve heard that one before. Our tracker has option based on sub-pixel accuracy that help compensate for noise, the default settings work great, but if you need to really fight the grain then bumping up the sub-pixel accuracy is definitely the way to go.”

PFTrack actually has a very comprehensive noise reduction algorithm called de-noise for noise reduction. These are specially designed algorithms that act similar to a median filter but maintain all edge detail. In essence they can be likened to high quality grain reduction tools.

Some trackers respond to image size. For example, some 3D trackers will produce a better camera track for film resolution material by actually tracking a video size resolution image instead. This is not so in 2D, says Discreet’s Schioler. “There is no reason to scale down the image. Under most circumstances, you should simply use larger tracker boxes. Image resolution doesn’t really change anything as far as motion blur is concerned. The important thing here is tracker box size.”

3D Tracking

Author Steven D. Katz, in the November 2001 issue of Millimeter, explains perhaps better than anyone how 3D tracking works. Match moving, optics, photogrammetry, and perspective drawing are all part of an area of mathematics called projective geometry. Applied to various spatial problems, it can provide solutions for measuring objects at a distance, locating objects in space, and extracting 3D models from photographs. Projective geometry has been used since the dawn of cinema in the form of glass matte paintings and false perspective sets. Camera angle projection, which is the process of extracting a perspective view from architectural plans and elevations, was the standard method used to visualize specific camera angles and views based on blueprints of a set.

Katz explains that “camera-matching software utilizes a subset of projective geometry called epipolar geometry. This branch of mathematics is used to describe the geometric relationship between two optical systems viewing the same subject and can be used to locate points in space. Because a moving camera offers a new view every frame, epipolar geometry works for a single moving camera as well, and each new view is understood as a separate optical system.”

The goal of a 3D tracking program is to solve the camera position by using two camera views. “Think triangulation,” writes Katz. Epipolar geometry is a type of triangulation. When using photographs to determine the position of a point on an object using triangulation, it’s necessary to match the image location point in one image to the image location point in a second image. Matching these pair of points in two images is called correspondence. Finding these matches would appear to require searching the entire image. Epipolar geometry proposes that the point we are interested is actually constrained to a single line. This greatly limits the search for that point.” By using the path of the point to define a line, one can predict where a point is going and greatly reduce the amount of computation and hence speed up the tracking process.

What is the ideal tracking marker to use on set for 3D Tracking?

Like 2D trackers, 3D trackers rarely if ever use colour information. “Chroma is relatively unimportant to boujou’s feature tracker,” says 2d3’s Steve Hill. “Pre-processing the images rarely improves the tracking results and we generally discourage pre-processing of any kind because of the many adverse effects that it cause. The exceptions are reducing the image size to reduce the effects of motion blur, and tweaking the linearisation settings of very dark cineon images to bring out more detail in the shadows.”

small
Click for Larger Image

Unlike the 2D experts we spoke with, the 3D experts have very strong opinions on what will track best and provide a large number of innovative tips on how to improve obtaining 3D camera solutions. Perhaps the most definitve comments on trackers comes from 2d3’s Hill who strongly advocates triangular trackers. “The advantage to the triangle in the circle is that it gives you three well-spaced, high contrast corners for boujou’s automatic tracker to pick up. The crash test marker has a high contrast area in the centre, and lower contrast areas around the edge. When this marker gets small and pixelated the central region starts to give four noisy tracks instead of one stable one (the detected features get pushed away from the corners by the blurring of the image at lower scales). The triangle in the circle works much better throughout a wider range of scales. Boujou’s target tracker works well on both sorts of marker, but for best results place the crosshairs of your keyframes in the centre of an area of black or white rather than at a corner.”

Digital Domain’s Doug Noble points out that their approach is very much that of modelling the environment from a highly accurate recording of on-set measurements. They aim to not so much give the 3D artist a camera track set of data points, but rather to “track the room to the camera” so the artist has a set with the camera moving through it. Noble’s TRACK software now usese optical flow and a range of solutions including Laser scans or “poor man’s LIDAR” (LIght Detection And Ranging).

3D tracking falls into two classes, automatic trackers and manual trackers. Many programs do both. Sci-d-vis’ Rolf Schneider, notes that “3D Equalizer has 3 different (manual) tracking modes: pattern, marker and edge/corner. The marker tracking mode is specialized on unicolored, flat discs placed on an unicolored background. 3D Equalizer computes the ‘center of mass’ of the disc which becomes the tracking point, on a frame by frame basis. That means that no previous frames are needed during the tracking process so no error accumulation happens. Marker tracking is very precise, but specialized. The cross/cheque patterns can be tracked quite well by regular ‘pattern’ trackers, that’s the reason why people like to use them.”

small
Click for Larger Image

3D trackers normally require a set of 8 to 11 points to be valid at any one time, but the same points do not need to remain valid for the entire clip. Unlike 2D tracking, automatic point generation is common in 3D tracking. This allows 3D camera tracks to solve seemingly impossible tracks. Perhaps the one of the most impressive and difficult tracks is to track helicopter moves over vast featureless oceans. This is extremely tricky as Hill explains, “Tracking water is always a gamble – if the sea is too rough then boujou won’t be able to find enough consistent tracks.” That being said, results can be outstanding as evidenced by the fact that Boujou has been used to track water movies such as ‘Enemy at the Gates’ (Double Negative) and ‘Troy’ (Framestore-CFC and MPC).

small
Click for Larger Image

Tom King, 3D artist at Digital Pictures in Sydney comments that he finds that Maya� Live will still track some scenes other automatic trackers can’t. He also feels it can also give more accurate Z depth than other packages at times. Maya� Live, like some 2D trackers such as Digital Fusion, show percentage confidence levels for the trackers in the UI. King suggests the following tips for Maya� Live:

  • Mmake sure tracking confidence is 70 to 80 percent green for all points. Bi-directional tracking can be repeated from a lower confidence area that looks right in the point centred window. Repeating this will result in longer high confidence sections.
  • Delete keyframes for sections that deteriorate into yellow and red confidence levels.
  • Using smart update behaviour in the options section can improve the chances of good tracks with points that change shape through rotation or lighting.
  • Make sure the Ready to solve graph is all green for the frames you want to track.
  • Before solving, point blast all point tracks to check for mistakes that are not detected by the tracker. One of these can completely ruin the Solve.
  • Do the solving in the manual stages and look at the pixel slip numbers as you go. If you aren’t getting a good solve, pointblast the tracking points again and correct any wandering points.

Dealing with Lens Distortion

One major consideration with 3D camera tracking is lens distortion. Lens distortion generally has very little influence on the 2D tracking process. However, it can have a substantial influence on the 3D camera calculation process. For this reason it is wise to film a grid with each prime lens (or in worse case film a known angular shape like a building). Most 3D camera packages have tools to adjust for lens distortion. For instance, 3DE includes a special lens distortion tool called “warpdistort”. It allows the user to remove and apply lens distortion to image sequences and can be useful throughout the entire production pipeline.

Lens distortion only affects the camera solving, not the pattern tracking itself. REALVIZ MatchMover experts Luc Robert and Ronald Mallet comment, “In MatchMover, lens distortion is represented by a mathematical model, whose parameters are computed automatically based on the point tracks. This model handles most of the lenses with a high level of accuracy. MatchMover also exports distortion nodes to Shake. In some cases, in particular when dealing with high-res footage, lens distortion effects are too important and too complex to be represented by a simple mathematical model. In this case, distortion can be learnt using a calibration pattern, and images warped before tracking to work on distortion-free footage. Later this year, we will provide specific tools to perform this.” 2d3’s new verison of boujou handles lens distortion but according to Hill, many facilities prefer to take distortion out of the footage using Shake, Inferno or other products before giving it to boujou. “If you go down this route,” Hill states, “you should make sure you use the right pixel aspect ratio for the custom camera that you will have to create in boujou. We’re going to supply a Shake node with boujou3 that will allow you to use the boujou lens distortion model to add or remove distortion. “. Pixel Farm, which also produces a Shake, Digital Fusion and AE plugin, can deal with high and low order radial distortions as well as distortions that change over a number of frames.

Many operators fear trying to solve a 3D camera move that has had a static resize in telecine. While a wider or larger picture will perhaps allow tracking markers to be valid longer before moving off screen, static resizes in telecine are not as disastrous as sometimes thought. Hill explains that while an accurate film back is important, it is not “as important as the pixel aspect ratio. If you get the filmback wrong then you’re going to get an inaccurate focal length value, but this is rarely a major problem unless the value calculated is way out. Pixel aspect ratio, on the other hand, can kill a camera track. It acts like a kind of 1D lens distortion and will have a very big effect on tracking accuracy. If the telecine process puts any squeeze on the frames then you’re in trouble”. If the image is adjujsted in telecine, it is important to know where the principal point (the projection of the camera optical center on the film back) lies in the image. With cropped footage, “it can even lie outside of the image,” jokes the RealViz team. “An error of a few pixels on the principal point will generally not affect the tracking too badly. Knowing the value of the camera focal length is helpful, both for accuracy and speed. Measurements on the set are also helpful. Any such measurement can be used by MatchMover to find a quicker and more accurate solution.”

The more powerful 3D packages also allow for additional information or images to be added to the solution. 3D-Equalizer allows stills to be added, perhaps taken by a digital camera from the side of set. MatchMover allows stills as well as data issued from ImageModeler. “This is particularly useful when dealing with multiple shots of the same scene, because it guarantees that all the solves are compatible and part of the same 3D world coordinate system. It also makes tracking a breeze,” they point out.

3D-Equalizer points out that all points should not be on the same plane, the 3D solver needs different z depth, this is a problem when markers are placed on a back blue screen. Here the problem is solved as – while the points are all on the same plane- they are different distances from the camera Here the issue is addressed by shooting at an angle to the blue screen wall.

The future of 3D tracking will most likely be linked to optical flow technology. In optical flow solutions, every point is tracked and it produces a set of floating point data which can then be combined with traditional 3D tracking for far reaching solutions. One of the leading companies in this new field is England’s PixelFarm, who appear to be leading the pack in multiple application uses of a core data solution. The 2D features are used to calculate the 3D match move, and those same 2D features can also be used to calculate the optical flow. This gives a very reliable solution to the optical flow calculation. PixelFarm’s Michael Lancaster points out that they “can also use ‘user features’ (manually tracked points) to guide the flow computation. So for instance, with old problem of a hand passing a face, stick a couple of user features on the hand and you’ll create a very much improved flow calculation. You can also use splines (mattes) in a similar way or in combination to define edges.” The problem PixelFarm address is a common issue for other tracking programs. 3D tracking is really a camera solution, it is not fundamentally about tracking things in the scene. In factm programs such as boujou provide matte generation to hold out moving objects that will interfere with the camera track. While programs such as 3D-Equalizer can be used to track people’s faces or hands, it is not their primary purpose. PixelFarm is building out from a core of tracking data to provide object tracking (including motion capture), camera tracking, motion blur analysis and even 3D scene reconstruction.

Prices Coming Down, Quality and Features Up

The other trend moving forward is dramatic price reduction. REALVIZ has introduced their ‘Absolu’ scheme in June of this year. ‘Absolu’ is a new monthly subscription scheme, which enables users who require occasional matchmoving facilities, to access MatchMover Pro regardless of their size or budget. PixelFarm has introduced PFMatch at only 600 pounds (approx $1000US). 2d3 released boujou bullet, a much cheaper ‘wizard-based” product with nearly all the functionality of the much more expensive boujou2. Andersson Technologies is selling SynthEyes for only $349US.

The price reductions are deceiving, because there is still an amazing range of functionality. One particularly difficult problem for 3D camera trackers are zoom lenses. While boujou users are waiting until September for zoom lens solving tools in boujou 3, other packages such as 3D-Equalizer, SynthEyes and PFTrack already include it. Shots taken from a tripod do not permit a full 3D solution, but SynthEyes can solve them to determine the pan, tilt, and field of view of the camera. SynthEyes also determines the 3D direction to each tracker, resulting in a spherical shell of trackers (panning, tilting, and zooming) surrounding the camera within the user’s 3D package. 3D elements can be inserted into this environment more easily than using a small number of independent trackers in a conventional 2D compositing application. But perhaps SynthEyes’ most amazing feature is its speed. Just as Benoit Sevigny did at Discreet years before, Andersson has focused on solution speed as a key enabling aspect of tracking and has been designed from the ground up to be fast.

“There’s no doubt in my mind that tracking will continue to evolve in a direction that relies on characterizing the entire image as much as possible, as opposed to specific feature-based analysis,” says Ron Brinkman. “Ultimately if your starting point includes information about the camera (derived from 3D analysis) then any additional tracking methods will have their accuracy improved dramatically.”

Summary of 2D Tracking Tools

Adobe

After Effects has recently improved its 2D trackers and provides a great set of standard 4 point tracking. While perhaps not quite as powerful as some of the bigger boxes, it remains one of the most-widely used effects solutions in the world.

Avid

Avid is the 800-pound gorilla of highend nonlinear video editing. In the area of effects products their tools are now headed by Avid|DS. Its Avid|DS and Nitro visual effects tools allow advanced compositing and tracking. DS Nitris is a standard and high definition finishing tool which offers power similar to After Effects and Photoshop. Nitro has a powerful 3D DVE, vector paint, motion tracking, keying, compositing, title and Symphony-like color correction tools. Nitro has come a long way since its relatively helpless title tool in the original Media Composer, but Avid still lags in marketing to the effects community. For example, Avid was the only equipment supplier who declined to contribute to this story.

Discreet – Advanced Systems

Discreet’s Advanced Systems, which include Flint, Flame, Inferno, Fire, and Smoke, all have 2D tracking. inferno and flame have 3D tracking. The Discreeet trackers are fast and very powerful and only lack multi-point solutions. Version 6 of inferno and flame contain an updated 3D tracker with much improved algorithims as well as auto-tracking. The lack of Maya export and a reliance on fbx as an interchange format is annoying and buggy. The 2D trackers are still the benchmark which other 2D trackers are compared to.

Discreet – Desktop

In 1997, Discreet aquired Paint and Effect from Denim Software. Paint offered a vector based painting and cloning system for Mac and PC, while Effect offered compositing capabilities. Discreet redesigned the interfaces to make the applications more Discreet like, and merged the two applications into Combustion. Along the way, they also replaced some of the core functionality such as Keying, Color Correction, and Tracking with the same tool set found in Discreet’s Advanced Systems. This provided flame-level tracking at a much much lower price. Combustion tracking files can be opened directly in the larger inferno/flame/flint/smoke/fire products.

Curious gFx Pro
gFx is a relatively new product for Mac OSX and Windows. Unlike other paint programs it is designed around a stong user interface that fully embraces moving footage, as such it can import, composite, track, or stablise footage easily. One of Curious’s founders is the man behind Parallax and Matador, which developed tracking on a PC back in 1995, and it shows in some of the depth of tools already available in the software.

Digital Fusion

Digital Fusion started in Sydney and moved to Toronto, Canada. At one stage, a version of Fusion was provided with Alias 3D. Today, eyeon has gained one of the strongest postions in NT/Windows desktop compositing solutions. Eyeon has two main products: Digital Fusion and DFX . Since Shake’s move away from NT/Windows Digital Fusion has provided a powerful cost effective solution.

Digital Fusion 4 is eyeon’s flagship product and marks the ninth major release of this powerful compositor. DFX 4 is the 8-bit expandable version of eyeon�s image processing software, Digital Fusion. DFX is based on the architecture of DF4 and offers a number of significant enhancements to its predecessor, DFX, including the flexible flow, superior character generation, PSD import into separate layers for animation, and more. Digital Fusions’s 2D trackers are some of the most advanced, and well integrated into every part of the product.

Bauhaus was founded in 2003. Bauhaus Mirage is a unified environment that simplifies the creation of animated graphics and special effects, including cartoon style animation in a 2 1/2 D environment. It features combined real-time video paint, animation, and effects functionality. Mirage runs on Microsoft WinXP/Win 2000 and Apple OS/X. Mirage version 1.2 followed version 1.01 for the PC, and was the first full OSX version. There have been some relatively small (hidden) changes in the tracking function with a view to smoother paths. Mirage’s tracker can work in a RGBA, or consider just single color channel, as well as either Luma or Saturation. Mirage’s tracker panel provides a monochrome magnified image of the the tracking area, along with a channel menu. This can be useful to visually judge which channel is optimal for tracking pruposes in a given situation.

Quantel


The Quantel product range consists of the, iQ DI, eQ post, and gQ broadcast graphics. All share the same code base and toolset including tracking. Quantel has made great advances in recent years in digital intermediate grading. Quantel has no 3D tracking but a comprehensive set of 2D tools in all products.

Shake

Apple’s Shake has a powerful set of 2D tracking software, but unlike some other compositing packages, Shake is only a 2D compositing package without a 3D camera view on the composite. Given Shake’s popularity this has caused many of the other specialist products to code special 2D export functions especially for Shake. This allows 3D tracking of complex camera moves and exporting special Shake data for tasks such as 4 point camera tracking.

Summary of 3D Tracking Tools
3D-Equalizer

One of the original 3D tracking tools, 3D-Equalizer is both highly powerful and widely used. 3D-Equalizer is not aimed at the dv pro consumer market. It is very much a high end feature film or major production tool. As such it has a serious learning curve but the web site has one of the most comphensive interactive teaching aids available of any company. The more data (focal length, survey points) which is recorded on set, the more information can be fed into 3D-Equalizer. This helps to reconstruct proper camera moves. Problems can arise when these data are wrong because of “human error”. 3D-Equalizer becomes confusedand it is difficult to debug by the user 3D-Equalizer is now avaliable for OSX and NT.

boujou and boujou bullet

Due to its relatively simple user interface and low stress operation, boujou has a very wide acceptance. boujou will track many things better than most, but until version 3 it will lack zoom lenses and some of the secondary camera footage options found in other applications. boujou bullet is even more user friendly. boujou and boujou bullet run on both NT and OSX.

monet

Monet is a 4 point tracker that actually uses planes rather than corner points. Properties of the source material such as lens distortion, shadows, highlights and grain/noise can be automatically applied to the new element. Combined with options for per-pixel motion blur, focus blur and colour adjustment controls makes Monet an extemely power specialist application. Monet was developed specifically for Cinesite in London for tracking the 2D animated paintings into the picture frames in Harry Potter III. Monet run on Windows, Linux and OSX.

REALVIZ MatchMover
REALVIZ is an outstanding company with a wide range of tools and there is a strong relationship between the technology embedded in all the REALVIZ products. For instance, MatchMover, ImageModeler and Stitcher share estimation algorithms that compute point and camera parameters based on correspondences across images. ReTimer is a little different since there is no 3D involved in the estimation process.

MatchMover has had a very powerful and flexible engine. It computes varying zoom as well as distortion and handles multiple shots and secondary helper pictures. The product incorporates both completely automatic tracking and supervised tracking, without limit in combining both approches. MatchMover can be adapted to any shooting situation: availability or not of survey data, information on camera lens, set constraints, motion control data, camera motion constraints (rail, tripod, fixed camera), different sequences from multiple viewpoints, 3D model-based tracking. Whatever data is available, you can take advantage of it within MatchMaker Pro. This versatility also extends to pricing through the Absolu pricing scheme.
Image courtesy of Jeff Mottle � CGarchitect.com

ras_track

Hammerhead sells their award winning tracking software directly. While numerous licences have been sold, it is not widely used in comparison to other leading packages. ras_track only runs on NT there are no plans for an OSX version.

TRACK

Digital Domain’s Academy award winning tracking program, now at version 5. TRACK is not yet available commercially outside Digital Domain, but Digital Domain has established a software division called D2 software to commericalise Nuke, Digital Domain’s compositing program. Nuke provides compehensive 2D tracking and stablisation.

The PixelFarm

The PixelFarm is a very exciting company, pushing perhaps faster than anyone else. The PixelFarm’s PFTrack is at the core of The PixelFarm’s data centric workflow. No longer purely a match move application, PFTrack is a means to extract complex 3D and 2D data from footage which can then be used in many ways. The PixelFarm provides a number of applets such as PFStable, PFRetime, PFPlate and PFBarrel to utilise data from PFTrack to enable operations to be carried out in a host compositing or editing application. In addition to match move tools, PFTrack now includes full use of survey points, advanced Motion Capture functionality, very comprehensive optical flow analysis tools and colour keying for easy masking of green/blue screen shots.

PFMatch is a low cost fully-featured match move application featuring the same industry leading tracking algorithms as PFTrack. Aimed at the single desktop user PFMatch offers a low cost solution for markets such as architecture or video users who require occasional match moving facilities and do not require the rich feature set and cross facility production tools of PFTrack. PFMatch hasmore tracking features than many of the higher cost tracking applications. PFMatch stands out against the competition due to it’s ability to use manual, automatic, or a combination of the tracking methods as well as solve all camera motions including zooms.

Both products can be used to pass calibrated data from single or multiple images and moving footage to PFBarn, a simple to use image-modeling application for film pre-vis and set reconstruction.

Maya Live
Alias’ 3D camera tracking program. It has not been developed much more in recent years but it is widely known for producing accurate z depth solutions. Maya live is packaged as part of the Maya complete package and is now available for both NT and OSX.