(Page 2 of 2 pages for this article  <  1 2)

Filed under: 3D •Audio •Cameras •compression •Hardware •Post Production •

HPA Tech Retreat 2010 - Day 4

Hugo Gaggioni & Yasuhiko Mikami, Sony - 4K End-To-End

New imager sampling structure: We have stripe-mask (F35), 3 sensors with a prism; both capture 100% in R, G, and B. Also Bayer mask; 1/3 as much data as full RGB, but 33% the resolution. How about Q67: Quincunx 67%?

Advantage of fewer chroma moire artifacts.

CMOS characteristics: jellocam from rolling shutter; various sensor-specific fixed-pattern and random noises requiring in-camera fixes.

Full RGB sampling is desirable, and Q67 gets close.

CMOS needs work but is getting closer to CCD.

High speed storage: SR memory, 5 Gbit/sec:

Hugo Gaggioni looks on as Yasuhiko Mikami shows off an SR solid-state module.

SRW-9000 will get SR solid-state recording, 35mm sized imager with PL mount.

Codec developments for HD, 2K, 4K: SR uses MPEG-4 simple studio profile (SStP)

HDCAM SR tape machines use levels 1-3, higher requires solid-state storage. New generation ASIC offers HD to 4K, all the way to uncompressed, adds SR Lite @ 220 Mbit/sec (SR already has 880 Mbit/sec HQ and 440 Mbit/sec SQ). Lite is designed for high quality TV without too much color grading & manipulation.

Moving towards editable, native SR files, wrapped as DPX or MXF. SRW-5800/2 (portable SR tape recorder) will offer MXF and DPX outputs, 220/440/880 Mbit/sec recording.

There will be a software-only player; for compression there will be a PCIe codec card for 2x ingest.

ELLCAMI project: universal transcoder with either Intel or Cell based motherboard, up to 16 Cell CPUs, 3.2 TeraFLOPS. “Anything-to-anything”, including new DEEP (Dynamic Element Expansion Protocol) upconversion tech running on Cell CPUS: SD to HD, 2K to 4K, blur-free, jaggy-free, etc. Uses wavelet analysis, synthesizes high-frequency texture elements.

4K projection is here, but not much content yet. Higher res requires orchestrated development of optics, electronics, sensors, etc.; standardization is key (SMPTE, AMPAS, ASC, other manufacturers).

Qs: Will DEEP be available for other CPUs, other vendors? “We’ll talk.” SStP open? Yes, SStP is an open standard. More Q67 data? Don’t even have a sensor yet: may be a good subject for HPA next year!

Aseem Agarwala, Adobe - Three-Dimensional Video Stabilization

Project motivation is tracking shots like the opening of “Touch of Evil”. Walking is a hard thing to stabilize: traditional stabilizing filters smooth it out, but it doesn’t look like a tracking shot. Most 2D stabilizers estimate motion, then fit a full-frame warp (homography, rotate/scale, etc.) that minimizes the jitter. But homography can’t model parallax, doesn’t know actual 3D motion path.

“Traditional” 3D stabilization derives 3D structure from motion, creates 3D point cloud, then plots a new camera path through the space and renders a new scene, usually combining/averaging nearby frames. But this winds up with jumps, judders.

Adobe’s improvement requires 1:1 input:output, frame-for-frame correspondence. Segment out layers, determine depth, shift and re-compose, fill holes. But can’t be totally accurate; insufficient data for complete accuracy.

Can we cheat? Aim for perceptual cleanliness, not complete accuracy. Use “as rigid as possible” deformations, image retargeting (as in Photoshop CS4’s image resizing without element distortion).

Take input point cloud for a frame and the output point cloud, perform content-preserving warping with least-squares minimization (trying to stay close to similarity transforms: uniform scaling, translation, rotation).

These content preserving warps can fake small viewpoint shifts, and the human visual system is surprisingly tolerant of remaining errors. But it’s time-intensive, a bit brittle, requires enough parallax to see/derive 3D geometry (so long shots without a lot of motion are troublesome), and rolling shutter is problematic.

New work (not yet published): subspace stabilization. How do we get these 2D displacements better? Take all tracked 2D points, low pass filter them? No good: trajectories don’t track, so image gets very distorted.

Is there something more lightweight than full 3D reconstruction? Track some feature points, put into big matrix; matrix should be of low rank. Factor the matrix into eigen-trajectories, plan smooth motion through those trajectories, remultiply it back out, get a new low rank matrix with smooth motion. Isn’t baffled by lack of parallax or a rolling shutter.

The demo videos screened show dramatic reduction in jellocam artifacts, and looked superbly smooth. Adobe is onto something here.

Works on streams; doesn’t need the entire content in memory at once (but does require low-pass motion filtering is not all content is available at once).

Some of the Qs:

• 3D? Multiple cameras make stabilization easier, due to more info. Haven’t tried on stereo yet, but will.

• New algorithm is sparse-matrix algebra? Parallelizable? Yes, but that part’s very fast already; the hard part is point tracking.

Steve Lampen, Belden - The Joy of Wire (Post-Retreat treat)

Mr. Lampen gave an amusing and inspiring talk about being open to new ideas and asking the right questions, ranging from the Aztec game of Tlachtli (which used a rubber ball; when Europeans saw this, they neglected to ask how the rubber was cured, and so lost the secret until Goodyear re-created it… and rubber is used to insulate cables!) to the periodic table of the elements (stellar fusion progresses from lighter elements to heavier, stopping at iron; heavier elements require greater pressures, like supernovas. Just think: copper is used for cables! That copper cable was born in a supernova… just as the elements in you were).

Some quotes from his presentation:


“When you exhaust all possibilities, remember this: You haven’t.” - Thomas Edison

“All this trouble to learn, when all you have to do is remember.” - the Bhagavad-Gita

“Where there is no vision, the people perish.” - the Bible, Proverbs 28:19

“With our thoughts, we make the world.” - the Buddha.

Demo Room Wrap-Up

Pro-Cache portable LTO-4 offload station from Cache-A.

Inner workings of Spectra Logic T950 LTO-4 robotic tape library.

LTO-4 tape is the archiving format du jour. Portable devices like the Pro-Cache or LTO-4-capable wrangling stations like those from 1Beyond allow tape backup in the field, which makes production bonding companies happier, The tray-loading Spectra Logic robotic libraries offer massive capacities of storage with high spatial efficiency (tapes are stored in trays of eight or ten tapes each; the robot slides a tray out, picks a tape from it, and then slots the tape into a drive).

(A fellow I used to work with came from the tape-storage industry, and he described tape as a WORM system: write once, read maybe1 Still, what else are you going to do that has any better chances of eventual readability, aside from Disney’s successive-color film recordings? Banks and bonding companies like tape; maybe they’re just conservative sticks-in-the-mud, or maybe they know something…)

In a separate suite, Joe Kane (of “A Video Standard” fame, and the more recent Digital Video Essentials test discs) was quietly showing a reference-grade projector he developed with Samsung. It’s a $13,000 DLP system with 10 bit, 4:4:4 capability (not sure if this is via HDMI only or if it has dual-link HD-SDI), 1920x1080 resolution, 85% depth of modulation at the pixel level, and no noticeable geometric or chromatic aberrations. No rocket science, just careful selection and matching of the DLP engine, the lens, and the screen. He says it’s been available for about four years (!) but it seems to be a well-kept secret. The USA distributor is Enders & Associates.

There’s a whole lot more; I only really took pictures of things that worked well in photographs (and once you’ve seen a photo of one double-imaged 3D display, there’s not really much more to see). There’s a complete list of the demos in PDF form if you’re interested in researching things further.

Final Thoughts

3D was everywhere, but no one I spoke to was sure that 3D will stick around this time. Few people are enthusiastic about 3D as a storytelling device… but its ability to get customers into theaters (“bums on seats” in British parlance) is well documented. Time will tell if that continues to be the case after the novelty wears off… or if the desire for 3D really does extend into the home viewing environment.

Last year, the demo room was filled with glasses-free lenticular 3D displays to the same extent as it was with polarized or frame-sequential 3D systems using glasses. This year, no lenticular systems were to be found.

TV loudness issues continue to be a major problem, but at least it looks like both the intent to fix them and the tools to do so are now in evidence.

Panasonic kept fighting off the perception that their new 3D camcorder will be consumer/prosumer kit. Yet when I talked to various Panasonic folks, I was told that the industrial design (looks like an HMC150, even though it has HMC40-derived guts) is intentional. My feeling based on this, and on the tone of Mr. Bergeron’s talk, is that Panasonic is purposefully making the camera look like prosumer gear, all protestations to the contrary aside. That way, they can afford to have it flop, or at least turn out to be too limited (if it looked like a binocular Varicam, the perceived stakes would be much higher). If, on the other hand, the little camcorder is a huge hit, Panasonic can follow it up with a version 2 unit with a fully-pro look and feel (though it’ll still likely use 1/4” or 1/3” sensors, just to keep size / weight / depth-of-field / interaxial under control). Just my own speculation, that’s all.

Yuri Neyman’s dynamic range talk on Thursday certainly ruffled a few feathers. There’s been a fair amount of email traffic about it and more than a few grumpy comments. We’re all waiting for Mr. Neyman to post his presentation on the HPA website (sorry, presos are in an attendee-only directory) so we can sift through it and suss why his figures are so radically different from everyone else’s.

As always, the Tech Retreat is great place to learn about issues of the day, and see where things are headed… but if you’re looking for firm answers to settled questions, you’re better off at a trade show like NAB, where the vendors claim to have solved all your problems (grin). The Tech Retreat is like a practically-oriented academic conference: you may not find exact solutions, but you will understand the problems better, and you’ll get your head stuffed full of exciting and thought-provoking ideas.

More:

Tech Retreat Day 1.

Tech Retreat Day 2.

Tech Retreat Day 3.

Other Tech Retreat coverage.

Demo room demo descriptions.

Recommended reading: Robida, The 20th Century (Robida invented the idea of newscasting, according to one of Mark Schubin’s fiendish quizzes).

16 CFR Part 255 Disclosure

I attended the HPA Tech Retreat on a press pass, which saved me the registration fee. I paid for my own transport, meals, and hotel. The past two years I paid full price for attending the Tech Retreat (it hadn’t occurred to me to ask for a press pass); I feel it was money well spent.

No material connection exists between myself and the Hollywood Post Alliance; aside from the press pass, HPA has not influenced me with any compensation to encourage favorable coverage.

(Page 2 of 2 pages for this article  <  1 2)

Adam,

Awesome as always. There is SOOOOOO much meat in this article that I had to re-read sections to just start to grasp it. And you posted all of these articles (days) so close together when, in blog parlance, you could have separated each person or topic and made each its own post and had more and better content - for months - than the majority of blog articles out there.

Thank you for feeding the inquisitive minds not with answers but with more questions and ideas surrounding the issues many of us are just beginning to grasp at the production level. For it is with deeper questions that we arrive at the right answers.

And thanks for being far above the:
- here was something interesting - [link]
type of author.
You are worth your weight in gold.

Anthony Burokas, http://IEBA.com
Publisher: http://TechThoughts.org

Posted by IEBA  on  02/24  at  09:14 AM

---

Name:

Email:

Location:

URL:

Smileys

Remember my personal information

Notify me of follow-up comments?

Submit the word you see below: