News Stories

DCS Notes – Day 1 – Session 1 – Understanding Stereopsis and 3D Image Capture

Session 1: Understanding Stereopsis and 3D Image Capture

Speaker(s):

Peter Lude, Senior Technology Executive, Sony Electronics

Steve Schklair, CEO, 3ality Digital Systems LLC

Peter Lude

Monocular depth cues (such as motion parallax, depth from motion, and perspective) contribute to our mind’s interpretation of 3D in the real world as well as stereoscopic 3D content. In a given image, something out of focus is perceived as being behind something that is in focus.  Vanishing point, or the convergence of lines as they approach the horizon, also provide visual depth queues.    Color intensity and contrast contribute; things that are far away are duller.  If you look at a 2D image with only one eye before you have seen it with both eyes, you will perceive it as 3D.  Only when binocular vision kicks in do you instantly snap into perceiving it as only 2D.

Mean interocular distance is about 65mm, with a wide variation.  Children start off with a distance 10-15 mm smaller on overage.

Positive parallax corresponds to seeing the right image on the right, and the left image on the left.  This places the object behind the screen plain.  Negative parallax crosses the eyes by positioning the right-eye’s image to the left of the left-eye’s image.  This places the virtual object in front of the screen, in ‘negative space.’

Mistakes in shooting 3D include; vertical misalignment of the image, non-synchronous lens zooms, mismatched focus, color mismatch, and keystoning.  The content should be authored for the largest expected display size.  If you author for a small screen and display it on a large screen, you will produce disparity – you are forcing the audiences’ eyes to turn outward in opposite directions.

3D camera rigs can either be two physically separate cameras locked on a bar, or a camera-pair looking through a silvered mirror beam splitter.  The beam splitter  creates a more ‘human’ interoccular distance.

Steve Schklair

Steve Scklair used a live feed during his presentation to illustrate shooting points and errors.  Production challenges include; developing the pool of trained crew, choosing the appropriate technology, revising the production pipeline, understanding the production budget and logistics, and developing the new language.

Two basic rig types.  The beam splitter simulates the interoccular distance.  It allows you to bring objects right up close to the camera.  Side-by-side rigs match the functionality of the beam splitter rig at a lower cost because no beam-splitter optics.  It is often used in sports because there is no reason to bring anything close to the lens.

Steve conducted a live demonstration of vertical misalignment.  Everything about camera-pair positioning must be remotely controlled, because you cannot have people running up to the rig with wrenches during a shoot.  The idea of keeping vertical locked in a zoom was considered critical from day one, because you have to zoom in when shooting sports.  Focus mismatch and zoom mismatch can occur even when you turn the zoom rigs identically, because the mechanics and lens characteristics aren’t identical.  This can be fixable in post, but it can cause discomfort during live events.

Too narrow an interaxial distance reduces the 3D to 2D.  It is ok to have sustained images 1-2% of the screen width in front of the screen, but holding images in front of the screen further and longer would be problematic.  Keeping the depth fairly consistent among the cameras makes cutting more comfortable, both for the audience and the editor.

A fix for edge violations is to focus on the closest object. Or you can just eliminate the edge violation by reconverging your cameras and putting the object completely in the frame.

3DIQ: Sky paid 3ality to put rigs into Telegenic trucks.  The lessons 3ality and Sky learned from the experience include:

  • • Editorial pace is slower from shot to shot (because there is more info in the shot)
  • • Staying a bit wider works
  • • It is important to be consistent with the depth
  • • It is important to level the depth across the edits
  • • For live broadcasting, fewer camera positions are needed in 3D than in 2D
  • • The story is more important than the WOW factor

On set, the monitors are good enough to view the shots as long as you are positioned properly.  Realignment in post will kill your budget.


DCS Notes – Day 2 – Session 9 – 3D 10 years Hence: 20/20 in 2020?

Session 9: 3D 10 years Hence: 20/20 in 2020?

Hans Hoffman, Programe Manager EBUI Technical

Panel

Tibar Balogh, cEO/Founder Holografika Kft

Dr. Aljoscha Smolic, Disney Research

Dr. Takayki Ito, NHK

Hans Hofman intro: There will be no interoperability issue in the future, because there will be no glasses.

Tibor Balogh, Holografika Kft

Home TV is in many ways a social experience.  You watch with your family or friends with the lights on.  You may talk about the show or something else.  You are not just staring at the screen.

To see the perfect 3D TV, look out the window.  When you cannot tell if it is a window or a TV then we have the perfect 3D TV.

Pure 3D will reconstruct the light field.  The light field is a function of position and direction.  Variants are X, Y, and angle for field of depth, plus vertical and horizontal parallax.  Reducing the vertical parallax can simplify the solution.

HoloVizio system – Optical modules project light beams to hit the points of the screen with multiple beams under various angles of incidence.  Holographic screen has a direction selective property.  It will be comfortable, glasses-free, no viewing restrictions, 2D/3D compatibility, and won’t cause any visual discomfort or restrict viewers.

The smallest practical pixel size is 0.1mm or 1 minute of arc, at 25cm optimal viewing distance.

There is a difference between trying to construct the 3D scene (light field) and tryig to mimic the human experience (view field).

Aljoscha Smolic (Disney Research, Zurich)

He described basic concepts of acquisition, transmission encoding formats, compression, delivery formats (BD, internet file), decode and regeneration/rendering, display.

With advanced 3D displays, Holografika today shows 60 or more views.  Other autostereo displays use 8,9,16 views without glasses.  But transmission of multiview data is very inefficient.  A spatio-temporal video continuum may be a better way to reduce/transmit/reconstruct the data.  He proposed reducing X views down to a few views accompanied by the data to reconstruct the X views.  He showed graphics imaging how this could be done.  MPEG will issue a Call for Propose in Oct. 2010 for specifications.

He is also involved in the MUSCADE initiative (www.muscade.com ?), which is a large European consortium.

Depth estimation of the 3D scene reconstruction must be accurate, robust, automatic, and real-time.  The form factor and handling of the acquisition systems is a problem.  The rigs must be manageable.

Stereoscopic Warping: Disney has developed a technique to manipulate the 3D without manipulating the image itself.  Good for corrections and conversions.  (His description was not clear.)  He will be presenting a paper on this at Siggraph.

smolic@disneyresearch.com

Takayki Ito (NHK)

Phases of NHK research

– Phase 1: basic research

– Phase 2: 3D TV

– Phase 3: visual comfort / discomfort, visual fatigue

– Phase 4: holographic

The four key natural visual cues are vergence, binocular disparity, motion parallax, accommodation.  The more natural cues you use, the better the 3D effect will be.

The cause of visual fatigue is inconsistent vergence and accommodation (i.e. the vergence accommodation conflict).

Test: He studied subject responses to: 2D still image, 3D still image with large disparity, 3D image with time varying small disparity, 3D image with time varying large disparity.  People experienced more fatigue when viewing time varying disparity than when viewing a 3D still image for a long time.

NHK has produced an integral 3D image which causes focus to vary with viewing z position!   The resolution of their prototype, shown at 2009 NAB, is too low.  Their research and development effort is dependent on technology advances.

< PREVIOUS ARTICLES NEXT ARTICLES >

Specification for Naming VFX Image Sequences Released

ETC’s VFX Working Group has published a specification for best practices naming image sequences such as plates and comps. File naming is an essential tool for organizing the multitude of frames that are inputs and outputs from the VFX process. Prior to the publication of this specification, each organization had its own naming scheme, requiring custom processes for each partner, which often resulted in confusion and miscommunication.

The new ETC@USC specification focuses primarily on sequences of individual images. The initial use case was VFX plates, typically delivered as OpenEXR or DPX files. However, the team soon realized that the same naming conventions can apply to virtually any image sequence. Consequently, the specification was written to handle a wide array of assets and use cases.

To ensure all requirements are represented, the working group included over 2 dozen participants representing studios, VFX houses, tool creators, creatives and others.  The ETC@USC also worked closely with MovieLabs to ensure that the specification could be integrated as part of their 2030 Vision.

A key design criteria for this specification is compatibility with existing practices.  Chair of the VFX working group, Horst Sarubin of Universal Pictures, said: “Our studio is committed to being at the forefront of designing best industry practices to modernize and simplify workflows, and we believe this white paper succeeded in building a new foundation for tools to transfer files in the most efficient manner.”

This specification is compatible with other initiatives such as the Visual Effects Society (VES) Transfer Specifications. “We wanted to make it as seamless as possible for everyone to adopt this specification,” said working group co-chair and ETC@USC’s Erik Weaver. “To ensure all perspectives were represented we created a team of industry experts familiar with the handling of these materials and collaborated with a number of industry groups.”

“Collaboration between MovieLabs and important industry groups like the ETC is critical to implementing the 2030 Vision,” said Craig Seidel, SVP of MovieLabs. “This specification is a key step in defining the foundations for better software-defined workflows. We look forward to continued partnership with the ETC on implementing other critical elements of the 2030 Vision.”

The specification is available online for anyone to use.

Oops, something went wrong.