The Foundry announced today the launch of Ocula 4, the latest version of its software plug-in toolset for NUKE, designed to address the problems with stereoscopic imagery. Ocula 4 introduces a complete overhaul of The Foundry’s industry standard stereo post production toolset.

Ocula 4 has been under development for several years and builds on feedback from the Ocula community.

New Disparity Generator

Ocula has always been very good at matching pixel detail between left and right eyes.  One of the problems in Ocula 3 was the clarity of the disparity map, in other words the edge clarity was lacking. This leads to foreground bleeding.

Ocula 3
Ocula 3 with the old disparity vectors accuracy
Ocula 4
Ocula 4 produces a sharper, thus better map

With Ocula 3 this edge softening would result in the artist having to do quite a lot of paint and roto. With the new Ocula 4 there is a whole new vector engine which produces a disparity map with dramatically less leakage and more clarity of depth separation. This failure to produce a sharp disparity map leads to an error called disparity ‘roll off’.The disparity generator is key as most of the main of the tools inside Ocula are built on the disparity map and more accurately the disparity vectors generated by it. The disparity is the difference between the left eye and the right eye. It shows the parallax and thus depth of a shot.

Of course, the whole point of stereo is the two eyes look and are different. Ocula helps users limit that difference to things that will enhance the stereo experience not distract from it. For example, one wants the left and right eye horizontally offset, but not vertically offset, from each other. One wants the images to be matching in color and not buzz or flicker. Images need to be sharp and accurate, but again with the same level of detail for each eye. Ocula 4 refines and extends many of the tools that have made Ocula the 'go to' product for stereo correction, - but some of these corrections are extremely complex and involve tremendous image processing.

For The Foundry, the approach has always been to make the best tools possible, but it has also been to acknowledged that in some cases there is no perfect one button solution in the real world. For Ocula, - as with many key image processing tools, - the mandate has been to make the tools as flexible as possible as these tools are being used on some of the complex stereo projects being shot. Thus, even if the tools can not perfectly solve a problem, users are happy if the tools can do the heavy lifting and get you 80% of the way home. 80% is very valuable and can make massive improvements to productivity.

In this respect Ocula 4 can be seen as maybe lifting that bar from 80 % to closer to 90%. On projects as labour intensive as stereo shows, Ocula also provides GPU acceleration, based on the intensive GPU Blink engine R&D Bruno Nicoletti started back in 2011. The disparity map generation is computational expensive so it really benefits from acceleration and the Ocula team have worked hard to make sure the GPU results match the CPU results.

The second big algorithm change, beyond the disparity map generator, is the re-writing of the image building tool that would be deployed by multiple ‘fix’ tools to generate a better stereo matching frame. Ocula’s famous party trick from the first day's it was launched was to move the left eye and the right eye, so both cameras actually look the same. In reality, a user will use the disparity map and the various Ocula tools to do a bunch of fixes from color, to gain to vertical alignment. What could happen in the earlier versions of Ocula is that that lead to a softening of the image. It is very helpful to autoalign the left and right eye images for color matching, - but what one really wants to achieve is image processing one of the eye’s to match the other, without softening, or killing the noise / grain or losing subtle high frequency detail in hair and cloth materials.

Inside Ocula 4 is a new engine that produces a much better image match. For example the O_NewView node produces a much better new pass when one needs to fully generate say - a left eye image from just the right. Not only will the image be sharper and match but it wont have tearing and a mismatch of depth at the edges.

Andy Whitmore, Chief Product Officer, says: “Ocula 4 steps up to meet the ever increasing demands of artists for quality results.. the new features include quick and easy alignment correction and automation, which sees the simplification of Solver controls, updated feature matching and the ability to auto-analyze sequences.”

One of the hugely successful tools inside Ocula has been the color matching tools to rebuild the color of one eye to the other. This detailed color correction can be very hard to do by eye. As an example of the 80% only solution approach mentioned above -  in Ocula 3 a shot might have succeeded matching the color of a sharp foreground character but failed to match the colors in the background. Helpful, - but it would still require roto and manual artist help to finish the shot, particular in the defocused background area.

With Ocula 4, we have more controls, we can use smaller block sizes for more accurate controls but with a ‘scale’ control that will limit the update to make it safe and tune the output to make sure it is inside tolerances you might set for the shot or sequence. The result is better, but the tools also allow more control to take the shot further before a user has to drop down to manual labour.

Tim Baier, is a very well known stereographer internationally and known to fxguide readers from articles as early as 2008. Baier consulted to the Dr Jon Starck and the stereo R&D team at the Foundry (UK) for the new version of Ocula. Tim Baier has worked on such films as Legend of the Guardians and at companies such as Animal Logic and Weta Digital. Baier has worked to provide day to day tool enhancements inside Ocula that are focused on the complex problems he has seen from year’s of well intentioned but mis-matched stereo footage. Baier ran test footage through Ocula 4 from films such as Hugo, Gatsby, Avatar and many others. The aim here is to provide very good QC tools that are very much battle tested at the time of release. Those who know Baier know he is a stickler for precision, workflow and practical implementations.

The StereoReviewGizmo is an example which provides a huge range of ways of comparing two matching images, with simple sliders and Baier’s elegant complexity is easy to see and use. Nodes such as the new VerticalAligner are workhorses designed to punch through shots and prepare them for a vfx pipeline. Much of this work also enable more automation inside Ocula for the reality of large shot count preparation and QC work in the real world.

The solver in Ocula4 has also been simplified while deploying more image processing and analysis than in Ocula 3. In Ocula 4 there is a new user match workflow, it is easy to set up half a dozen user matches to lock down the alignment on a plate with a new assisted GUI.

Finally there is a new stereo retimer, which produces new matched clips with less jello effect, again all because of the new disparity vector generator and image generation algorithm.

Avatar, used with permission. Copyright.

Thanks so much for reading our article.

We've been a free service since 1999 and now rely on the generous contributions of readers like you. If you'd like to help support our work, please join the hundreds of others and become an fxinsider member.

  • Pingback: Ocula 4: Exclusive info about the new release | Occupy VFX!()

  • Henrik Cednert

    I do wonder how much of Ocula’s underlaying vector generator that’s shared with Kronos? Let’s hope that the work on Ocula will ripple through to Kronos so that we’ll see some improvements there as well. =)