3D LOST AND FOUND
Daniel L. Symmes
This article is dynamic in that it will change over time as new or
different information is found.
As I have been studying, collecting, and making 3D films since…well, a very long time, when my friend Jeff Joseph (www.sabucat.com) sent me an email announcing he had found a “lost” 3D film, I was more than interested.
There are a number of significant 3D films missing, and likely lost.
One title missing in action, A DAY IN THE COUNTRY (DITC), was a short released in March of 1953, one month before the big 3D explosion with the simultaneous openings of two 3D features, MAN IN THE DARK and HOUSE OF WAX.
Running about 13 minutes, it was produced and directed by Jack Reiger and distributed by Lippert Pictures.
On first sight, I felt the film was shot in New Jersey. Instinct, I surmised.
While working on it, I found a major clue. During a runaway car scene along the country roads, there’s one POV shot approaching a stop sign. Behind the stop sign, across the intersection, a sign that is both out of focus and blurred by camera movement. Yet, I found a couple of frames where I felt I could make out “New York 62” (or 58 or 43) and “Sussex 3,” among others.
My friend Google Earth allowed me to find Sussex, NJ, is in fact about 63 miles from New York. However, there are different ways to measure such distances.
But Google allowed me to find what I believe is the intersection in the film, just down the road from Sussex (3 miles, to be exact).
The subject matter, silent with narration by comedian Joe Besser, is the story of two boys and a grandmother (?) visiting a farm in the country. Poorly produced except by 1915 standards, but if one looks close, the humor is decidedly warped.
The short was originally exhibited in the anaglyph 3D method, using blue/red glasses.
3D historians pine over thoughts of seeing “the rare, the lost.” So Jeff’s find was highly significant.
While I was restoring the film, many questions cropped up. Jeff Joseph, and associates Bob Furmanek and Jack Theakston were all over it. “What 3D camera did they use?” being the number one question.
What we knew was that this short was picked up for distribution by Lippert Pictures (very low budget production/distribution) in JAN53. That would mean it was shot late ’52, or perhaps minutes before it was acquired.
After viewing it a number of times, I proposed a camera system that could fit the bill, but I had technical reservations. This film exhibited a fixed interaxial, and no use of convergence. Virtually any of the known 3D systems of time wouldn’t have these restrictions.
But I still felt it likely to be a known system and until something better came along…
... it did.
In the first week of OCT06, Jeff Joseph amazingly found paperwork that the original title was STEREO-LAFFS, and that it was "licensed" for exhibition in NYC in 1945! Amazing, and answers why the film looked older—it was!
And most recently, Jeff found a reference to it being available early 1941, indicating it was likely shot in 1940.
But now the mystery of the 3D system is even more so. As all people involved are deceased, it seems we likely will never know.
The last theatrical 3D at the time in the U.S. was the MAR41 MGM release of Pete Smith's short,THIRD DIMENSIONAL MURDER. In watching the stlye of DITC, it is quite apparent Reiger used TDM as his model.
Currently, it is believed the negatives and even other prints are not likely to exist. The Lippert collection was fragmented many years ago, and most of the shorts have disappeared. So this print is VERY rare —likely the only one extant.
Anaglyph films (and print images) are based on color to separate the left and right, superimposed, stereoscopic views. The colors should be complementary (opposite) and various combinations have been tested over the years. The dominant combination is cyan/red. As “cyan” covers a wide range of colors to the eye, it can look blue to a distinct shade of green. Also known as "blue-green."
The best cyan viewing lenses are those that allow an even distribution of blue and green (simply, minus red).
The red must not pass ANY other color before the “green/orange” boundary at a wavelength of about 580nm (minus green and blue).
The big problem with anaglyph is that the colors of the print aren’t good enough. With color print films the red image is created by the film’s magenta and yellow layers, and can be said to be a synthetic red. The cyan layer is not a spectrally “clean” cyan. Thus, the red and cyan colors used for anaglyph are not perfect.
If you view an anaglyph film image and see “edges,” or “ghosting” (crosstalk) then (assuming the glasses are correct), the dyes are not right.
All anaglyphs have this problem to some degree. So when one approaches an anaglyph film for film restoration, the results have in the past been far short of successful.
This is because the first level of restoration is to capture the anaglyph as it is and hopefully improve on it. Since the original is essentially defective, any film-based duplication will not only bring its own degradation (contrast, color inaccuracy), but will invariably exacerbate the color impurity of the original.
So for DITC it was decided to go direct to digital.
Jeff and I have an on-going relationship with International Video Conversions (IVC, www.ivchd.com) in Burbank. They are one of (if not THE) leading HD labs and general manager Scott Call has been more than up to the many challenges we’ve brought his way—he almost dares us to bring him impossible problems. Almost.
As DITC looks like a film made in the 1910’s quality-wise, we decided to do a standard HD transfer (versus 2K). The Spirit DataCine recording to D5, and under the experienced control of colorist Mary Kay Miller, was all we needed.
Phase one was to make an “archival” transfer attempting to obtain an anaglyph equal to the original or better—a master copy of the original, warts and all. Regardless of our ambitions, we always want a version that represents the way the film WAS for subsequent academic study.
This is largely a dial-twirling exercise with one eye on the vectorscope and the other looking through appropriate 3D glasses.
NOTE: To view the following 3D images, make sure the RED lens
FIG 1 - Video transfer of the original anaglyph. Notice the extreme
parallax (image doubling). Color was faded making color-based extraction difficult.
From the start, this wasn’t apparently going to work well since the film was shot without the benefit of a 3D setting called convergence. To the naked eye, everything is VERY double. For properly shot 3D films employing convergence, the images can look almost normal to the eye.
This extreme parallax, as it is known, means one’s eyes get exercised too much, and makes restoration difficult.
We managed to increase the chroma and massage the luminance quite a bit, and ended up with a better anaglyph than on the original print.
The second phase is to make left and right eye separations.
The anaglyph was originally created by taking a left and right “stereo pair” (views or eye views), coloring one cyan (sometimes green), the other red, and superimposing them making one image. In a perfect world we would have access to the original L&R elements. Sadly, we were incredibly lucky to only have a faded release print.
The goal was to “extract” the original left and right views.
The way to extract is by “keying” on the color in question (cyan or red), in the hopes of “seeing” only one of the two images. Since there are imperfections, even the elaborate electronic systems with secondary color manipulation can’t do a perfect job.
But, with a lot of elaborate and sometimes counter-intuitive manipulations of every kind available to Mary (plus some against-the-rules board-pulling in the Spirit), we managed to get ONE eye (right) nearly perfect, and one less-than-perfect.
The important thing is to get one CLEAN view if possible as will be discussed.
Both views are recorded as B&W images. And because they originate as colored images on the film, the grayscale images we derive are not matched in an acceptable manner. If we try to do more than a global adjustment, the time needed would adversely affect the budget. As this effort was not truly “funded,” we are at the mercy of little money and kind hearts.
even with elaborate electronic adjustments. On the right the Right eye separation (red image).
Very little crosstalk, though a too dense.
Phase three, is to bring the three versions to DIMENSION 3 (D3).
Since about 1980 I have been developing techniques and procedures to take existing 3D films and make corrections to make them better 3D presentations.
These corrections are rather complex and sometimes interactive; changing one often affects one or more of the others, usually negatively.
I have a Prime Directive that I will not change the artistic INTENT of material I work on—these are other people’s dreams. What I do is correct as much as possible problems in the original photography including convergence (or lack of it), camera alignment, focal length variations, and rotational errors.
Any one of these aspects can cause eye discomfort to the point of being unwatchable.
I can guarantee this short was unwatchable in 1953. Yet, a review written at the time was quite enthusiastic. I truly believe this is because the author hadn’t seen better 3D. Without a decent reference, even the worst 3D (eyestrain-wise) can be “acceptable” to the average person.
The excess parallax also distracts from this otherwise entertaining 3D “demo.”
While transferring DITC, Jeff asked how I would get rid of the ghosting on the one eye.
My recollection is an immediate reply of “you could take the negative of the good image and apply it to the positive of the bad image to subtract the unwanted aspect of the bad eye” (the left eye in this case).
I was taken aback by this spontaneous concept.
The next day I did some experiments to establish basic settings and parameters and was encouraged.
DITC’s “good” view is the right. It had very little ghosting. The left view had many issues.
I brought both views into parallel time lines in Adobe After Effects. The right eye was inverted to negative, and combined with the left view. I could then interactively adjust the right view until I “removed” a maximum of ghosting from the left view. This was exciting to see working.
In use, I found there is no one setting that worked for all scenes. It depends on the shot: exposure and contrast are key variables. If the views were identically exposed, the job would be much easier. But these separations were very different in visual characteristics, mainly because of the anaglyph coloring of the print.
Then I could see another defect. When I effectively subtracted the right view from the left, there was then visible an edging around the residual “ghost.” Apparently object transitions (edges) have an aberration that might be caused by the original silver image or the color dye. Might be a diffraction issue.
For example, a person’s head, with black hair, against a lighter background appears to have a thin, bright white line, and another adjacent black (or gray) line. The white line is usually brighter than the scene, so it’s often quite visible.
FIG 3 - After subtracting, a residual “edging” or outline of the subtracted image is visible.
The arrows point to the dark haired boy’s white outline to the left of his head.
Not obvious in many shots, but in a high contrast image (this film is very high contrast),
it is noticeable to technically-inclined observers.
Video, particularly composite video, can have “edging” or boundary artifacts, especially at the interface of contrasting colors. This isn't a video issue, but is similar in principle.
So while my subtraction process works well, it makes this outline more prominent. Yet, even with this issue, the subtraction works so well that the overall improvement of the left view is still much more acceptable.
It is my belief this outline defect can eventually be isolated and removed. But as this restoration was handled on a “no budget” basis, we’ll have to work on this issue in the future.
Once I have a “corrected” left view, it is possible for some shots to do the reverse where there’s ghosting in the right view, using the corrected left, subtracted from the right. But this is possible for only a few shots overall.
Most crosstalk removed and scene grayscale improved.
With the resultant “clean” L&R views, it was then possible to make an “archival” anaglyph that in most cases is superior to the film original. In fact, it likely is better than the original print was in 1953 since there’s much more control over the cyan/red colors of video.
This archival version is made to preserve what WAS, albeit better than it was.
I then found that several cuts in the short were 3D reversed. That is, the cameraman or editor mistakenly switched the left for the right. So in 1953, not only was the 3D difficult to view because of the excess parallax, some scenes were 3D reversed. Yikes!
Again, staying within my Prime Directive, I corrected the reversed shots. They made an error, I simply corrected it.
Then the critical work began.
While working on a previous 3D restoration, THE STEWARDESSES, I decided to give my restoration efforts a name: 20/20 PROCESS™. That came from my seeing my work as hindsight: I can see what is wrong with the original, and can correct it with today’s understanding and technologies.
While I can correct for many problems, I cannot change the amount of 3D in a given shot. That is determined by the distance between the two, 3D camera lenses (the interaxial distance, to be technical). The wider the distance between the lenses, the greater the 3D. Many of today’s 3D films have too much 3D (parallax), from too much interaxial.
DITC had both too wide an interaxial AND lacked the use of convergence.
With any film, I start by superimposing the left and right views, and move one left/right and up/down to get to a point where I can determine two other aspects: sizing and rotation.
Most professional 3D cameras have well matched lenses. Some don’t. This means one image is slightly larger or smaller than the other. I correct this by reducing the larger view to match. Always reduce—never enlarge.
Rotation deals with one view being rotated relative to the other. This is rather easily corrected.
Once sizing and rotation are dealt with, I then level the images (up/down), and study the convergence (or in this case, lack of) on a shot-by-shot basis.
Convergence is, in this environment, simply the left/right alignment of the images. I can therefore easily place the convergence anywhere I choose.
If the film is well shot, such as HOUSE OF WAX, the convergence is left largely alone. But for most films, at least some convergence adjustments will improve the 3D viewing experience.
Dynamic convergence, which changes during the shot, is possible but I go to lengths to avoid this. Some people have proposed this silliness for production these days, sadly. Yes, there is an occasional shot that NEEDS this, but for the most part it should be avoided as it CAN be visible, and detract from the presentation.
Another adjustment at this stage can be the overall color/exposure matching between the views. As normal policy, I rarely do this unless requested by a client—my motto being “you get back what you give me” as far as the image grading is concerned.
I had to adjust DITC as the two views poorly matched. Brightness, contrast, apparent gamma. There was no way the images would ever “match.” But human 3D perception is very adaptable, and even major imbalances can be compensated for mentally.
Lastly, as I had to swap the left for the right views for the reversed footage, matching was that much more difficult.
Admittedly, if I had many weeks to work on it, I could have achieved fractionally better results. There’s never enough time. Yet, I feel the results are better than expected, even by my standards.
The 20/20 L&R views can be used for dual polarized projection, and the making of other 3D versions. The 20/20 anaglyph is MUCH easier on the eyes than the original, making the film viewable for the first time.
Anaglyph 3D has gotten a bad rap from journalists. You know, the “goofy glasses” syndrome.
True enough, it isn’t nearly as good as proper polarized 3D projection. However, polarized projection is a projection technology—it cannot be recorded nor broadcast on TV.
The ONLY process to distribute 3D films on DVD, cable, broadcast, and shown in ANY theater is anaglyph. From my perspective, this is the ONLY way to get 3D to the “masses.”
And, as the film was originally exhibited in anaglyph, we’re doing justice to history.
The bonus is that as we have L&R masters, this film can now be seen in polarized.
The 20/20 restored version was shown as dual projection, polarized 3D for the first time ever at the World 3-D Film Expo II (www.3dfilmfest.com), at the historic Grauman's Egyptian Theatre, on Hollywood Blvd., on the afternoon of 17SEP2006. The audience was obviously entertained.A DAY IN THE COUNTRY will never be highly regarded, though it is mildly demented.
I am indebted to the help and contributions of energy and simple time (our most important commodity) in allowing me to bring this rare “find” to the bright light of today to the following:
Jeff Joseph, Jack Theakston, Bob Furmanek, Scott Call, and Mary Kay Miller.
All content, unless otherwise indicated, is
© 2007, Daniel L. Symmes
All rights reserved.