Stereo3D Geometry

­Also see:
Stereo3D Clip Coupler | Stereo3D Ghost Toaster | Stereo3D Image Balance | Stereo3D Monitor Scaler
Stereo3D Parallax Guides | Stereo3D Text Generator | Virtual Floating Windows | Stereo3D Z-Depth Map

Introduction

New to the Stereo3D Toolbox suite is the Geometry plug-in, which provides a more comprehensive way to edit and manipulate all measures of Stereoscopic depth & geometric disparities. It allows for percentage-based corrections, which is crucial when working with proxies of full-resolution 2k & 4k footage as all corrections made will still have the same stereoscopic effects on the native footage. Any disparities in camera alignment at the point of acquisition can be corrected with the Geometry plug-in down to a sub-pixel level and be outputted and viewed in a wide variety of formats.

Initial Setup

The Geometry plug-in, as with all the plug-ins in the Stereo3D suite, can be applied to an already pre-existing multiplexed stereo pair (left & right video files married together) that is any number of 3D formats (side-by-side, over-under, etc) as well as be applied to a single left or right eye file and then be coupled with the correct adjacent video clip. For more information on coupling left & right files together with S3D Toolbox please refer to the Clip Coupler guide.

To access the Geometry plug-in go to:

Video Filters → Dashwood Stereo3D Toolbox → Stereo3D Geometry

Once applied to the clip, the clips render status will appear red in the FCP timeline unless set to Unlimited-RT. To access the menu to change the timeline’s render settings click on the RT button on the far left of the timeline. Select Unlimited-RT and the clips render status will turn orange, indicating that Final Cut Pro will process as much of the un-rendered effect in real-time as permitted by the video card. Typically the more VRAM that FCP has access to, the closer to real-time it will display the un-rendered clip. If an external 3D Monitor is being used for playback, then Playback Video Quality should be set to High and Playback Frame Rate can be left at Dynamic.

If the footage being used has already been muxed into a stereo pair, then the correct S3D Input Type must be chosen in order for Toolbox to correctly process the image. Refer to the examples below for clarification of the Stereoscopic formats.

3D- Indicates a format that will display images in Stereoscopic 3D.

Interlaced 1080i. (Row-Interleaved, Line by line)  3D

Interlaced will display the left eye data on the odd lines of video resolution and the right eye data on the even lines, which in turn works natively with most Passive Polarized 3D monitoring systems. Often referred to as Row-Interleaved, because of “interlaced” more often referring to how fields are being displayed rather then meaning a Stereoscopic format, Interlaced has more often become a display format than an acquisition format. S3D Toolbox allows for Interlaced 1080i as an Input Type, because if another S3D Toolbox plug-in from the suite was set to output Row-Interleaved, then if new video effects like Geometry was placed on top the video, then S3D Toolbox can interpret the stereoscopic interlacing and then display it in a new or similar format.

Side-by-side. (SBS, Left & Right Squeeze, L/R)  3D

One of the more commonly seen standards of the 3D formats is side-by-side, which can be natively received by most 3D Monitors and be output for stereoscopic viewing. The left & right video files are placed beside one another on the horizontal plane and are squeezed in by 50% in order to fit into a single frame. Side-by-side can also be not squeezed for two full-resolution frames in the same composition and video file, just twice as large in resolution.

Side-by-side Crossview (Cross-eyed)  3D

The exact same as the normal side-by-side except for that the eyes are swapped so that the viewer can cross their eyes and view the scene in 3D without glasses or a 3D display. This can cause great eyestrain, but with practice it can become a simple rudimentary way to “free-view” 3D material without the use of expensive equipment. The footage typically isn’t squeezed in crossview.

Over Under. (Above & Below)  3D

Also a common 3D format, the left video track is squeezed vertically and placed on the top 50% of the frame and the right track is then also squeezed vertically but placed on the bottom half of the frame. With S3D Toolbox, objects that lay converged or at Zero parallax are stitched back to full resolution when they are displayed in 3D.

2D Source

Stereo3D Toolbox allows for the input of a 2D object, such as text or a lower 3^rd to be manipulated in 3D Stereoscopic space. This does not mean that the object has been converted into true stereoscopic 3D but rather has simply been duplicated into both a left and right eye file so that the interaxial distance can be controlled. The object can be maneuvered and key framed in and out of negative and positive parallax or rather “in” and “out” of the screen. 2D source should only be applied to an object that does not take up the entire resolution of the composition as it is intended for an individual object within the scene and not a full resolution composition.

S3D Input

Once the footage has been loaded into the timeline with the S3D Geometry plug-in placed on the clip, the correct S3D input will need to be selected. Clicking on S3D Input Type will access the input drop down menu and the input can be changed by selecting any of the formats. If the footage hasn’t been “married” or muxed together yet, then any of the top 3 Image Well selections can be chosen. Footage that has been acquired natively or “baked” into Anaglyph and chroma based 3D formats are not suitable inputs.

Geometry Features

SbS is Horizontally Squeezed (Only available when using Side-by-Side input types)

This option tells Stereo3D Toolbox whether or not the side-by-side footage is squeezed into one full resolution composition or if it is two full resolution compositions coupled in the same file.

Swap Inputs

The naming hierarchy of left & right 3D files can sometimes be incorrect depending on the accuracy of the Data Management Technologist’s file handling and the labeling of the physical cameras on a Beam-Splitter Rig. This naming error will be evident if the footage is viewed on a 3D display, as the background will be perceived to be floating in front of objects that should be in the foreground and vice-versa. When incorrectly labeled the left and right eye video will be reversed and thus the human eye will be receiving incorrect information.  If the viewer is unsure if the footage is named properly, then by clicking Swap Input back and forth while playing the footage, the 3D will either be comfortable to watch or slightly straining to view. If the footage is comfortable to watch and the objects within the scene are positioned in the proper location in 3D space (foreground objects appearing at the front or “popping” off the screen & background objects appearing in the back or “behind” the screen) when Swap Input is selected, then this indicates that the footage has been named incorrectly.

Global Output Mode

Global Output Mode is a keen way to set and apply the same stereoscopic output mode to an entire sequence so that every clip does not have to be adjusted individually. Selecting the Use Global Output Mode check box on a specific clip ensures that that clip will adhere to whatever the current Global Output Mode is set to. To set the global output mode, select the appropriate output using the Global S3D Output Mode drop down menu, which only becomes available when Use Global Output Mode has been selected, then click the Set Global Output Mode icon.

This setting is particularly useful when a sequence needs to be edited using a viewing method such as Anaglyph and then exported out to a stereoscopic standard such as side-by-side or over-under. Ensuring that every clip in the timeline has the Use Global Output Mode checkbox selected will certify that the entire timeline will output to the correct format.

S3D Output Modes

The various stereoscopic 3D output modes available in Stereo3D Toolbox can be accessed underneath the S3D Output Mode Tab. Side-by-side, Row Interleaved and Above Below have been discussed in Initial Setup.

Checkerboard 3D

Similar to Interlaced in that it displays both the left and right eye video data simultaneously, Checkerboard is typically found in conjunction with the use of most 3D DLP viewing systems. Rather then alternating left and right eye data on a line-by-line sequence, checkerboard displays the left and right eye video streams on alternating pixels, shifting the starting order on every new line of resolution, thus making a checkerboard pattern. This format is being used less and less.

Column Interleaved 3D

The same premise as Row-Interleaved in that it sends a left image to one line of resolution and the right image to the next line except that it interleaves the data into vertical columns rather then horizontally. This format is rarely seen but has been known to be used on various European formats.

Superimpose (50/50)

Changing both the left and right eye video files to 50% opacity, superimpose allows for both videos to be seen in 2D simultaneously overlaid with the least amount of visual disturbance or artifacting. It looks similar to Row-Interleaved but is not a stereoscopic format and can work at any scaled size.

Difference

Used solely for checking the convergence of a scene, difference will indicate which object is converged when the pixels turn black and begin to disappear.

Anaglyph 3D

The most classic 3D format, anaglyph requires the use of glasses with tinted chroma based filters so each eye will see only the desired image, filtering the opposing eye’s image in order for the 3D effect to be seen. Stereo3D Toolbox allows for 4 different anaglyph formats including Red/Cyan (Simple) and Red/Cyan (Advanced) for more highly detailed adjustments and chroma correction incase there is too much red or cyan lighting within a scene. Blue/Yellow and Green/Magenta have also been included which are popular formats for many DVD releases of theatrical movies. Anaglyph is known to create headaches when viewed for prolonged lengths of time but is the only 3D format that can be universally used on any television set provided it has colour output and that the viewer attains the proper pair of Anaglyph glasses.

Convergence Controls

The convergence sliders are very powerful tools that are the bread and butter of all 3D plug-ins and software. To be able to control the various elements of convergence allows for the correction of footage that wasn’t properly aligned, altering of the interaxial-distance and the overall stereoscopic effect. Simply by sliding the Convergence faders back and forth will adjust the entire scenes placement within 3D space. The convergence controls can correct headaches that the audience might feel from watching misaligned 3D and if used incorrectly can also give the audience eyestrain and nausea.

Convergence Coarse allows for x-axis or horizontal adjustment of the left and right eye images that will change whether the scene is pushed further “off” the screen and into negative parallax or whether the scene will be forced back into the screen behind the screen-plane and into positive parallax. Each value of 1 equates to 1% of the screen width. Typically, if the footage was shot correctly, this value should never reach dip bellow -2 when outputting for a home television based release and the maximum is typically -1 for theatres (Refer to Parallax Guides for accurate values of the maximum divergence of a 3D composition based on the display screen size).

Convergence Fine performs the same as Coarse, but only on a sub-pixel value which is crucial for extreme accuracy and for adjusting the image for a theatrical release on a large format screen.

Adjustments that have a negative value (slider to the left) shift the left image to the left and right image to the right pushing the entire scenes stereoscopic 3D effect “back” into the screen. Whereas adjustments with a positive value (slider to the right) shift the left image to the right and the right image to the left pushing the entire scenes stereoscopic 3D effect “off” the screen and towards the audience. Any left and right images that lay perfectly overtop one another will appear on the screen plane or zero-parallax and be perceived by the audience to be located right on the screen neither “popping off” or going “into” the screen. Objects within the scene that are at zero-parallax will look as though they are a native 2D image without the aid of 3D glasses.

If the interocular distance between the cameras is too large for comfortable viewing, then a negative value adjustment should be introduced to correct the extreme convergence. By bringing the left and right images closer the 3D footage will be easier to watch and to enjoy. As with all sliders in Final Cut Pro, Convergence Coarse and Fine are fully key-frameable, providing a powerful way to change the 3D placement progressively throughout an entire scene.

Disparity Correction

The alignment of 3D camera rigs is finicky and sometimes is a painstaking process that can consume a great deal of time on set. It’s not uncommon for shots to come back from production with misalignment issues that will need to be corrected in post. Due to time constraints during production, it can sometimes be more feasible to adjust smaller disparities in post then rather on the day of, BUT it is always encouraged to try and achieve the most accurate alignment on set. Despite some of the best efforts during production, it is a guarantee that there will be misaligned shots present in the post process but Stereo3D Toolbox allows for full correction of these possible issues.

Besides horizontal disparity or rather convergence, left and right eye images should match each other as accurately as possible so that the viewers brain is receiving to similar images in size, colour and geometric position.

Vertical Coarse works on numerical values similar to Convergence Coarse except that each value of 1 equates to 1% of the screens height rather than width. It is very crucial to be aware of vertical disparities, as it can be very discomforting and painful for the viewer to watch them. Naturally the human eye sockets do not have any vertical disparity or misalignment and by displaying footage that possess discrepancy between the left and right eye vertical placement causes eyestrain and the ocular muscles above the eyes to flex and pull in an attempt to try and correct the images.

By switching the S3D output mode to Difference, vertical disparities will become more evident when there are bright pixels separating the top and bottoms of the left and right images. Horizontal lines amongst the scene are also a great way to examine for vertical disparities.

Moving the slider to the right on Vertical Coarse will push the left eye image up towards the top of the frame and the right eye image down towards the bottom. Moving the slider to the left will produce the opposite effect.

In Difference mode, the bright pixels will disappear as the vertical alignment becomes closer.

Vertical Fine allows for sub-pixel manipulation of the vertical alignment. All misalignment issues are unique and Vertical Fine adjustments are crucial to achieve the accuracy needed for ideal realignment.

Rotational Disparity will correct any rotational issues if the cameras have “rolled” during the shooting process. If a camera is tilted to either side it will create a rotational disparity that may not be evident near the centre of the frame but often intensifies in the far edges of the frame creating vertical and sometimes horizontal disparities. Moving the slider to the right will rotate the left image counterclockwise and the right image clockwise using the centre of the image as the anchor point. Moving the slider to the left will produce the opposite effect. The Difference Output Mode will once again be helpful to spot and correct rotational issues.

Zoom Disparity will correct discrepancies between the two cameras zoom lens focal lengths and the flange depth (the depth from the lens mount to the sensor). No two prime or zoom lenses are ever the exact same. These differences are sometimes very subtle and only detectible by highly calibrated Flange Depth Readers, where as at other times, when a zoom lens is set to the incorrect length, they can be very evident in the footage.  Zoom disparity can take sometimes take longer to finesse than other disparities as it can be difficult to match object proportion and sizes. Centering left and right images so that the objects converge overtop one another within the composition is an easier way to more accurately match the zoom and focal lengths. The images can be returned to the original positions after the rotational issues have been addressed.

Keystone is designed to adjust and correct the “keystoning” effect that occurs on 3D camera rigs that shoot converged rather then parallel. If shooting converged, there will be differences in perspective that can become evident in the corners of the composition. There will be slight geometric skewing and by using the Keystone adjustment, Stereo3D Toolbox will manipulate and rotate the images along the Z-Axis. This will reposition the compositions so that there is “corner-pinning” and no longer keystoning.

All changes can be saved as Disparity Presets and opened at a later time. This allows the user to create a database of templates for various camera rigs and shoot-dates so that the plug-in can be applied effortlessly to footage at any point to correct alignment.

Enable Autoscale

When Convergence and disparity corrections are made the composition can be left with black sections of pixels where the original left and right eye video files were before, before they had been shifted. For simplicity, the Autoscale checkbox allows for one-step automatic resizing so that the composition fills the entire frame removing any black pixels from the sides and corners.

Master Reframe

Master Reframe allows for re-composition of the shot while maintaining the 3D alignment. X% changes the horizontal location of the composition while Y% changes the vertical location. Master Rotation will rotate both the left and right eyes simultaneously in the same direction and Master Scale will change the size of the files.

Adjust Eyes Individually

Oftentimes only one eye will need adjustment and further finessing if only one of the cameras was out of alignment. As well, cameras shot on certain beam splitter rigs need one or both of the cameras to be flipped or flopped in order for the perspectives to match. By clicking the Adjust Eyes Individually check box these features will becomes available in the geometry plug-in to apply exclusive adjustments to each eye.

X% will refer to the horizontal location of the eye and in turn will affect the convergence.

Y% will change the vertical location of the individual eye and thus the vertical alignment.

Rotation will rotate the left or right image clockwise or counterclockwise incase only one camera had roll introduced.

Scale is the equivalent of the Zoom slider used for disparity correction.

Flip will rotate the image 180º so that it is upside down from it’s native position.

Flop will change the images perspective so that is a mirror image of it’s native position as though it has been rotated 180º along the z-axis.

Once again, all of these sliders can be key-framed for each eye incase the alignment progressed throughout a shot.

Output Options

Swap Outputs

It is sometimes necessary to swap the output of the clip for many reasons such as some 3D Monitors interleave their polarized signals differently line by line. If ever needed, the Swap Outputs checkbox can be selected after the convergence and stereoscopic settings have been changed.

Also see:
Stereo3D Clip Coupler | Stereo3D Ghost Toaster | Stereo3D Image Balance | Stereo3D Monitor Scaler
Stereo3D Parallax Guides | Stereo3D Text Generator | Virtual Floating Windows | Stereo3D Z-Depth Map