When small interaxials just don't cut it! - Page 32

Interpolation algorithms thrive on smooth continuous motion. Throw them some random motion, and the predicted motion becomes unpredictable.
Edited by MLXXX - 6/28/12 at 3:50pm

As a lover of 3D, I really enjoy scenes involving water droplets, or waves. (Or schools of small fish for that matter.)

My only reservation is that if the IA is very wide, some scenes can lose their impact. I don't imagine that would be an issue with the fountains at Bellagio with a 6" IA but it could be an issue if using a large IA to capture a choppy sea. The apparent height of the choppiness can be much reduced so that the sea looks more like a calm lake, in 3D!

Jeff- the math is pretty well defined in stereography. The stereo base used is a relative function of the stage size to put it simply. I have not yet calculated the range of stereobase for optimum 3D range for that stage but with wider angle lens the range of stereobase permitted is wider too. 6" was my best guess based on the memory of the nearest key fountain directly in front. When I do the actual computation, it may turn out to be optimum at 4" or 8" Going too wide can cause diverging disparities too great for eye comfort and I have learned my lesson at guessing now and won't be making that mistake again. I will use the stereo base calculators and the distance range finder for a better SB choice.

In your example, a choppy sea can look small and flat if it is of great distance and shooting wide angle. If you zoom in on the sea at great distance, then increase the SB you will get better z depth, not less. If you shoot too close and the SB is too wide, then the sea simple diverges and you see double images.

Don, I noticed unnaturally flat water in my own 3D still of the view from an ocean liner leaving Port Canaveral [using frames extracted from 2D video footage] ; and in a video of Frank's involving yachts/boats. As you say, if zooming in on water, the parameters would change.

A difficulty arises when the main subject is distant, requiring a higher IA, and there is at the same time water visible in the near foreground not far from the camera position. The water in the near foreground will look unnaturally flat because of the dwarfism effect introduced by the high IA.

I think the beauty and grandeur of fountains is in the water spraying through the air. I don't think a degree of dwarfism in the appearance of the ripple of waves in the pond into which the sprays of water drop would be of too much concern; though excessive divergence of LEFT ad RIGHT must of course always be avoided. [ As you say, calculations can avoid disappointment.]

In a best case scenario, the eagle would grab the stranded copter and drop it, gently, in front of the jury rig.Though pigs flying seems more likely.

I see what you try to test. Was not here for some days, but without testing I think following. If this idea is worthwile really, an interpolation should not be necessary.

Let me explain why I think so. The deinterlacing could take place simply by shrinking down the 1080 lines in a field to 540. So, if you take out every second line of a frame in an uff footage, you end up deinterlaced.

If we are successfull to do that for both fields, it would be great - since then you do not loose quality due to the deinterlacing and interpolation steps.

But that is my thinking. I do not know if it works out really.

Frank,

With your "eagle watching rig", what are you using for the wireless HDMI transmission?

VitrualDub can give a quick result, provided it can actually open the source video file (it prefers AVIs) . These internal filters could be used:

1 . To get half-height 30p from 60i:-

2. To get half-height 60p from 60i:-

I have used the MVC to AVI converter at http://www.3dtv.at/Products/MvcConverter/Editions_en.aspx to obtain decoded Left and Right AVI files (uncompressed, so very large) from Sony HDR-TD10 files.

Wolfgang, I was wondering whether you'd had a chance to look at my timing mismatch video from post #924, The Ungoverned Anniversary Clock, at http://www.youtube.com/watch?v=k_m4ETc-ydY

I'm seeking feedback on whether an optical illusion of the orbs appearing to fly further outwards for one direction of movement, and then further inwards for the reverse direction of movement (compared with watching in 2D) is perceived. (I see the illusion myself even at the minimum timing discrepancy of 4.17mS, beginning at 3m 15s into the video.)

Yes,such an approach would work out. My idea was to do that in Vegas, by rendering the L and R stream as separate files to 1920x540, with no deinterlacing. The field drop is the way how it has to be done, but I have not tested that yet really.


Sorry, I was so busy that I have not found the time to analyse that. I would love to do so, but time is pressing at the moment here.

No worries Wolfgang.

___


Actually I see that Vegas can be set to the following for rendering:
Video for Windows, 1920x1080, 29.97i, Upper field first, Uncompressed.

Perhaps Vegas would leave the interlacing intact for such rendering settings. On the assumption that that would be the case:-

If the project were set to LEFT, a short segment could be rendered as an AVI and that loaded into Virtualdub. Then the project setting could be changed to RIGHT and another AVI could be created.

VirtualDub could be set to produce full height bob-deinterlaced frames. These could then be loaded back into Vegas as full height LEFT and full height RIGHT and paired as a stereoscopic pair. [Or an AVISYNTH script could be used to combine half height source frames, of a clip 1 and clip 2, using the function: stackvertical(clip1, clip2).]

I'm not suggesting this as a usual work flow, but as a way of creating a short 3D test file of bob-deinterlaced video to see whether it really is an improvement over blended/interpolated.


VirtualDub is great for these types of exercises as you can step through the AVI frame by frame and inspect any detail closely. It can also produce a compressed version of the file pretty quickly (XviD MPEG-4).

I'm not retired (yet) and I don't have as much time as I'd like to experiment with some of these things myself.

Shot the Jacksonville Fireworks Wednesday. Just finished editing it.

This experiment was to shoot with the twin TD10's using 1080 60p ( Sony PS Mode) The stereo base was set at 18 inches and the distance to the barge on the River was measured to be 1239 yards. The near object in my shot was mostly too dark to detect but on occasion it was illuminated by the fireworks and was a boat at 340 yards. By these measurements I could have used my 1 meter slide bar to separate the two TD 10's. As such the z depth wasn't as good as I had hoped for but there is always next time. Last Winter I use a 9" and 12" Stereo Base. Anyway the event nearly got rained out as the thunderstorms gave us a break about 20 minutes before the show started.

The shooting and pairing went fine in Vegas (build 595) Playback could not keep up with the 60p ( Double NTSC) timeline and only played single clips at 25-28 FPS. This made my edit decisions a little more time consuming. The performance was supposed to have been synced to a music sound track that was to be played on the Riverbanks PA system but for some reason that never happened so I added my own music appropriate for the 4th of July event. ( Remember, this is for me and family as my target audience. ) Cost me $130 sync rights! Being retired from the TV business, I no longer have my blanket license for BMI and ASCAP. library so now I have to pay per song. So if any of you have a suggestion for cheap good quality music, let me know. There used to be a company with good prices on YT rights but they stopped doing it.

I think the biggest issue I observed working with 1080 60p is the rendering time. A video shot in 1080 24p on this length and effects added would take about 40 minutes to render but the 1080 60p takes about 2 hours! When I watched the video the motion smoothness didn't look obviously better than the 24p, but then my 24p presentation has always looked quite smooth on my monitors with the Sony a bit smoother than the Vizio. (240Hz vs 60 Hz refresh rate) Last night I rendered the show to an iso file at 1080 24p template in Vegas and the action looked very smooth so shooting the 60p and converting it to 24p for blue ray doesn't seem to be a problem. The onkly really good way I can see the native 1080 60p from the timeline is to set my Dynamic Ram to about 8Gb and then do a shift B render for ram playback. Unfortunately the time I get on this is only about 10 seconds!

I should have a compressed version of the Fireworks show uploaded by this evening on YT.

Don,

I don't know if you've checked out the Smartsound Library. They have fully customizable music that you can buy on a per song, per album or mutiple-album basis. It covers a wide range of genres, classical to hard rock, and music with vocals, too. I got 28 albums for $500 last year. I still haven't downloaded all 28 (I have a year to download the ones I want). They also have sound effects albums. I bought all 6 (100's of effects) for $100 recently. It's not for everyone, but a LOT of professionals use it. You can hear samples just by going to their web site. Their program is called Sonic Fire Pro, which is available for $100, and you get a couple of sampler albums with it. One really cool thing about it is that you can create custom audio-follows-video tracks. You're not limited to creating just stand-alone custom length music, although you can do that if you want to.

The Jacksonville Independence Day Fireworks is up now.

On my Vizio it looks best in 1080p using SBS half. LIne interleave looks fair and anaglyph is not good at all on mu Vizio. On the computer monitor in window it is better but some colors do not work well for anaglyph. The Wells Fargo bright amber often glows double.

The windows Media video I used for YT upload seems to soften the image too, compared to m2ts or MP4 rendering here. But the m2ts file size was 2.6 Gb.



Edited by Don Landis - 7/6/12 at 1:54pm

[Don, I think you should check the vertical alignment at around the 10 minute mark; and in the opening few seconds.]

Don, I always like the pacing of your videos!

I thought the orchestral music a nice accompaniment to the fireworks.

Regarding interaxial distance, I think the 18" you chose was the right distance, in order to be able to include the foreground water in the opening section of the video. (For my own vision, I think any more than that could have made viewing of the foreground water uncomfortable.)
Edited by MLXXX - 7/6/12 at 5:45pm

Thanks Jeff for your help on this. As I said in the PM and for others to learn from I have this current version uploaded with some auto correct errors. Starting with the opening shot, I accidentally unchecked the auto correct and then never rechecked it before rendering the video. Making the corrections took a few minutes but the rendering of this will be about 2 hours before uploading. I should have the new version up tomorrow morning.

Glad you liked the music. Had to pay some healthy coin ( on my budget) for the sync rights.

On the stereo base. I'm still trying to achieve some Fireworks negative parallax. I have only seen this done with those Fireworks simulation display programs and that isn't real. I want to create a fireworks star field in the center of my home theater in front of the screen. For the next local Fireworks, maybe Labor Day, I plan to use the 1 meter rail and then zoom in tight on the explosions. According to my SB calculator, the distances involved could work with a Hyper SB of up to 20 meters! Minimum distance can be no closer than 500 meters and the lens set for 135mm ( A little less than half the TD 10 range ) To do the water and wide shots, I can just slide the cameras closer together, then edit out the manipulation. Our shows seem to have plenty of opportunity where they have a break for about 10 seconds so that's when I can make the change in SB.

Don, I have downloaded the video as sbs half and have seen it on my 50-inch HDTV. Looks great so far.

My recommendation: I would tend to adjust the parallax at the level of the firework to a small figure. Especially a firework tends to ghosting, given the strong contrasts - so parallax should be low here really.

Yes I have been finding it useful to adjust parallax so that the main part of the subject of interest is nil or near to nil. It can make viewing more relaxing, as well as reducing ghosting.

Yes Don I think that would have to be the way to go to get dramatic 3D for the fireworks and still be able to include some wide angle shots of the surrounding terrain.

On the question of timing there may occasionally be some momentary mismatches between Left and Right with asynchronous capture because fireworks can produce a very short-lived but brilliant flash. Here's a side-by-side frame from Don''s original version of the Independence Day fireworks illustrating this (frame 375):



The Left frame includes a brilliant flash illuminating the water, and the Right frame does not. Frames 374 and 376 (Left and Right) both show the water in darkness. This was a short-lived flash!

It might be interesting to try to edit that flash into the right frame. It would no doubt require some careful work to get the angle right, but since it's so short it should be doable.

The sync between the left and right cameras was initially set with the Lanc Shepherd and was noted to be 1/120 in my shooting slate. The clapboard was then used after rolling cameras about 5 seconds and once again about 10 seconds before pausing cameras. The sound pulse on the edit wave form was lined up first and then during several places the the fireworks were checked left and right to see if the two clips were in sync between the cameras, about center of the recording. I saw a very close match up of the sparks pin points and felt there was no cause for concern. The sound and video misalignment on full expansion of the timeline was about 1/8 of a frame both in the beginning clap board and at the end. The distance to the Wells Fargo building was 1240 yards as measured by the Bushnell Range finder. This distance was assumed to be close to the barge where the fireworks were set off and also about same as the distance to secondary explosions high in the air. Sound travels at 1100 ft per second so I assume the missync of sound is about 3+ seconds. I did not slide the audio timelines with the video because I didn't feel the multiple explosions sync was worth the effort.

According to some stereographer experts I've read, when there is a missync of a strobe light flash between left and right cameras, the visual simply loses one of the eyes and loses 3D effect but if the light is sustained for both eyes to see the light then the visual will show the 3D illusion. If that statement is accurate and it seems to me to be so, then the moment of light first seen from an individual explosion particle would lose it's 3D but only for a tiny fraction of a frame exposure.
I think this topic is interesting from a theoretical angle but in reality, I'm not so interested in defining sync down to the nanosecond. If out by more than a frame I will correct. In my shooting, I'm happy with 1/3 frame accuracy. IMO, camera alignment is a much bigger issue to worry about with wide stereo base shooting.

BTW- the corrected upload is going on right now. When up, I will delete the first one with the errors.

New link will be:

From Don's updated version I made a loop of 100 frames with frame 375 in the middle of the loop [VirtualDub labels the first frame of a clip as zero, so this would be frame 376 in some other software]. This was to test whether frame 375 would "stick out" for my vision when repeatedly playing the loop: it didn't. So I think repair would be unnecessary, at least for me. The discrepancy was too brief to be noticeable. Also, this could have been a perfectly correct disparity if there had been an obstruction between the light source and the Right camera. The brain would need time to analyse that...

I did however notice "watery" effects occasionally and when I stopped the playback sure enough there was a mismatch between Left and Right in the part of the image that looked watery. It would not be the main part of the firework "constellation" that had just exploded into visibility but a "rogue star" that was moving downwards much faster than the other points of light in the constellation

But I am very sensitive to stereoscopic 3D. I suspect that most people wouldn't notice the small timing mismtach effect, or if they did it wouldn't bother them.

A few comments, having now seen the updated YouTube version:-

Enjoyable video. Really liked the music that went with it. Interesting to see the partly hidden fireworks going off behind the building on the left. They really did look "behind it" stereoscopically.

It's quite impressive what standard of pyrotechnics is available today: very complex patterns, precisely coordinated.

What more could we ask for? A wider IA?

Continued to make another major correction for a very subtle improvement after the latest upload. This improvement exists on my BluRay iso file rendering. All it did was deepen the blacks and sharpen the image slightly at tyhe moment of the biggest explosions. Here's what I did--
I noticed several times where I looked for them that the right camera seemed to have a brighter flash than the left. I think this was due to my manual exposure setting just a tad higher than the left camera. While I believe you may be interpreting this to be a timing problem the position of the point light was identical between left and right. Too close to change by sliding the left and rights by one frame length. Taking a different approach, I added two instances of Brightness / contrast effects where one was left view checked and the other was right view checked. Now I can adjust the brightness and contrast of each individually after pairing. That's all it took and I was able to match the brightness and the resulting glow of the explosions precisely between the left and right cameras. I made that correction after uploading the YT version but before rendering my Blu Ray version.

One of the problems we had that night was the heavy rain and thunderstorms for 3 hours just prior to the Fireworks. This caused a bit of ground fog over the river. Between that and the smoke, the diffusion of light was severe. Another serious photographer was next to me and had commented that the fog and heavy smoke patches lingering ruined most of his shots as compared to a clear night in January.

I think the only way to achieve the kind of sync perfection you seek Jeff is to use genlocked cameras or a frame store synchronizer like what Frank just bought. The Lanc Shepherd hopes to achieve sync between the two cameras by simply turning them both on as close to the same moment in time as by chance. So, in my opinion, any attempts to quibble over the intra frame timing mismatches with this Lanc technology is like trying to control the roll of the dice to get double sixes on every roll, because that is exactly how this Lanc system timing works. One keeps powering up over and over until the start times are closer than before, maybe get lucky and achieve perfection. At least in genlock you have a single clock to time both cameras. In frame store the equipment delays one video in reference to the other for electronic alignment and once matched, sends the video pair to the output.

Thanks, Frank. There is not one time, I am out shooting with this twin camera system that I don't say a thank you for talking me into this wide stereo base end of the hobby. It's not that I can use it to make the image more real to life, but that I can add to it for presentation.

Now there's an idea, pick and choose any pair camera set to see which stereo base looks the best.

Soon, if my plans work out, I will have two rigs, one with TD10's and one with a pair of wide angle NEX5n's. I do plan to set up all 4 on the bench to have two focal lengths to shoot these events next year. Then I can switch between the two in post. The TD10's will be able to zoom but the wide angle NEX5n's will be locked in 12mm wide angle. Both shooting at 1080 60p and all 4 synced with the clap board. How's that? Just pray the stock market keeps me funded so I can buy this stuff.