Tuning Challenges

IMR says their algorithm is highly tunable. Depending upon the bandwidth of the wireless technology employed, they could dial the compression back to, say, 75% instead of 95%, and at that point the difference in image quality may indeed be indiscernible even to the knowing eye.

I saw evidence of this algorithm tuning in action from when I first got a look at IMR’s compression tech back at CES 2017 earlier this year. Between then and now there was a marked increase in image fidelity, even at the same compression rate. The difference was so drastic that I went from feeling like the tradeoff between image quality and IMR’s wireless tech was not worth it (at CES), to feeling like it was good enough (during this latest visit) to warrant very serious consideration of whether or not to cut the cord on the Vive.

Photo by Road to VR

Of course, as with any compression technology, some things compress better than others. There’s still a few areas presenting challenges to IMR’s compression. Chief among them are gradients and dark/low-contrast scenes. Both can cause some amount of banding and artifacts which wouldn’t be seen otherwise. IMR says they have ways of dealing with those elements and are continuing to tweak their algorithm to compress different parts of the scene smartly to avoid artifacts. But you can imagine the challenge of getting the algorithm to behave correctly 100% of the time across lots of different VR content, each with its own optimal compression.

To aid the algorithm in ensuring the right compression approach for the content, IMR says that it would be possible for games to send parameters to the algorithm for per-game adjustments which could potentially allow for fine-tuning to be done by developers or crowdsourced to users.

Today & Tomorrow’s Headsets

A concept rendering of what IMR’s wireless VR add-on could look like | Photo courtesy IMR

The little white box I got to see was very much just a proof of concept. The next step for IMR is turn that box into a reference design that can actually be mounted on the head for truly wireless gameplay. The company is in the midst of creating the reference design and will soon demonstrate it publicly. At the outset it will look rather bulky, but IMR says that the system could be easily miniaturized and integrated directly into VR headsets of the future.

IMR says that their compression tech can support current generation VR headsets like the Rift and Vive using existing wireless tech like 802.11ac. For future headsets pushing higher resolutions, the company is hoping for the proliferation of more cutting edge 802.11ad tech which has the bandwidth for up to 4k per eye.

SEE ALSO
WiGig Chip Company Peraso Technologies Closes $42M Financing Round

According to the company, the algorithm has also been architected to be ready for eye-tracking, a feature expected to come to VR headsets in the near future. With eye-tracking data piped into the algorithm, IMR says it can further optimize its compression for the areas where the eye is looking, while saving data elsewhere (similar to foveated rendering).

– – — – –

With the demonstration of their core technology, IMR says they are seeking a partner who would manufacture an add-on device that would allow the Vive and Rift to be adapted for wireless VR, and that could come to market in 2017. The company is also sharing their reference design hardware with partners who may want to integrate the technology directly into a VR headset.

1
2

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. See here for more information.


  • Joan Villora Jofré

    Price?

    • Raphael

      19.95

      • Joan Villora Jofré

        For real? Where did you see it?

        • Raphael

          Henry told me. He’s a good source of information.

          • elev8d

            Jorge told me they were giving it away for free.

          • Raphael

            Is Jorge a reliable source?

          • elev8d

            He owns a mobile phone and has internets.

      • Joan Villora Jofré

        “Price point would depend on quantities but roughly around $1200-1500”
        http://www.roadtovr.com/imr-building-wireless-video-system-power-4k-per-eye-vr-headsets/2/

        • Raphael

          Consumers won’t pay $1200.

          • Maybe it’s not for consumers… yet :)

  • 1droidfan

    How could this work with Rift? The Rift sends audio and tracking information across the USB cable.

  • Seems very interesting. Image quality is reduced a bit, but if they continue improving the algorithm as they’ve done in these 5 months…

  • Timotheus

    It’s very interesting, that all the points and ideas that came to my head, where brought up later in the article. ^^
    At first, when it was mentioned, that no new high bandwidth wireless technology is used, but compression, I tought “Why not use as much bandwidth as possible and then invent an open new VR compression to compress the rest?” only to read, shortly after, that they do exactly that. Just their compression algorythm is proprietary.
    And then I thought, with eye tracking, even less bandwidth had to be used, only to read again, that they already thought about it and are prepared for it. XD

    So 1 ms sounds good. And given they fix some artifacts, it should be indiscernable from normal cabled quality. I hope they don’t just reduce quality in the peripheral area, when eye tracking is there and include NVIDIAs research, where they found out, how you can’t even notice, when the quality in the peripheral area is reduced.

    With some new wireless technology even 4k or 8k should be possible, especially with eye tracking.

    A Vive 2 with eye tracking, min. 4k and wireless in-built, next year, is on my wish list.
    Given, that the technology for every one of those technologies is already here, it should be possible.

  • What is the approx battery time? A stored charge that small on the headset strap would be what? an hour?

    • elev8d

      They haven’t gotten to that point yet. Both the transmitter and receiver were powered.

      • Daniel Fitzgerald

        Hi elev8d. This is not correct. As comments above, we have made a complete wireless reference design including integrated battery. Please see my links to the video above showing the IMR open night wireless demo.
        thanks
        Dr Daniel Fitzgerald
        CEO, IMR

        • elev8d

          Thanks for clarifying Daniel and providing links to videos. This just got more exciting in my book now. From the article it didn’t look like the product was that far along. It’s good to see it in action.

    • Tim Lucas

      We currently have a fully functioning head mounted prototype, wireless Video / USB with a self contained. quick swap battery that is tested and exceeds 2 hours of game play. CTO at IMR.

  • DaKangaroo

    “IMR claims their proprietary compression technology doesn’t use frame-to-frame compression (or sub-pixel sampling) and can achieve 95% compression with encoding and decoding happening in 1ms.”

    So what they have developed is an image compression algorithm, which they are just applying to each frame.

    I’d believe their claim of 95% compression because that’s pretty much in line with what you’d expect from a good image compression algorithm.

    For example, a 2160×1200 image is 2,592,000 pixels. Uncompressed, in 8bit RGB format (24bit colour, aka 3bytes per pixel), you’d be looking at 7,594KB of data. 95% compression would be 379.7KB.

    JPEG can easily achieve this with a low quality compression for example. A quick test in Photoshop and I was able to compress a wallpaper image that size with Photoshop (Quality set to 20 out of 100) and the results didn’t look too particularly bad when viewed at 100% zoom.

    JPEG is by no means the best algorithm, so I’m sure they could at least match that or do better. The tricky part would be figuring out a method that doesn’t require heavy computation. GPUs manage this for images by compressing images in blocks, see algorithms like DXT, that in some variants can achieve 1/6th image compression. Perhaps they created something along those lines.

    • Sounds like i-frames without motion-prediction (normally h.264 is I, B and P frames). Since they kind of implied they’re not using h.264 at all, it’s likely just HEVC run this way, though as you said, it could in theory be any sort of image compression. The trouble is that without B/P frames, bitrate is wasted on frames that don’t need it and not available for frames that do so you’re guaranteed with this approach to see visual degradation for extended periods of time (since ALL frames get chocked) IF they aren’t allowing for enough bandwidth.

      Of course, most VR games are quite simple visually right now (save for a few of the higher-end ones) so that critical bandwidth level might be low enough to not matter for those apps. For nicer ones, I’ll be interested in seeing the results.

  • Peter Hansen

    The white 3D printed box achieves perfect wireless transmission conditions in terms of distance, occlusion, reflections and so on. Would be very interesting to see how the device performs in real life conditions. Likely, with 802.11ac this is not much of an issue. 802.11ad is a completely different story, though (see the beam forming approach from the MIT which is supposed to be integrated with the TPCast).

    Would also be interesting to know how it compares to the TPCast in terms of latency, image quality, and signal stability.

    • Tim Lucas

      Hi Peter, I am IMR’s CTO, you are correct there are further considerations with AD when roaming, vs static. We have recently surpassed those hurdles and currently do have a fully operational, self contained, head mounted unit in a 4-player, multi-player setup.
      This works perfectly fine and allows for rapid movements and normally expected game play. I won’t make biased claims over competing solutions other than saying that our demo-able results speak for it :)

      • Joan Villora Jofré

        Hi, Tim:
        What price range would we be talking about for the current versions of HTC Vive or Oculus Rift (in dolars)? Less than 100, less than 300 or more? :)

        • Peter Hansen

          I believe I have read something about $1.200-1.500 in a prior article. For a system allowing up to 4 users with no considerable degradation in terms of image quality or latency I’d say this is a reasonable price.

          • Joan Villora Jofré

            Yes, in fact I also said it, but only for the connection? Nitero is talking about a system that costs less than the cable.

          • Peter Hansen

            Many people/companies are talking about many things, atm. It is important to see whether and what they are able to deliver.

            TPCast’s single user device, e g., comes for $250 pre-tax, their business multi-user model for up to 6 headsets will definitely be more expensive.

            Developing a compression technique independent of frames surely wasn’t cheap either. I think 1.5 k is fine for multi user – which is definitely not aimed at the home/consumer market.There will be other, cheaper products for that sector, but it remains to be seen at what quality.

            For me, IMR’s approach currently is the most promising of the 12 systems I have heard of so far, most of which are in a far earlier development stage.

          • Joan Villora Jofré

            If we want virtual reality to advance, it must be cheap; It can not be that a link costs like the PC needed or the HMD.

          • Peter Hansen

            Sorry to disagree, but it has to be well priced. And this is small business application.

          • Joan Villora Jofré

            Fortunately, it seems that the price has nothing to do with the rumored and is cheaper.

          • Peter Hansen

            Would love that!

        • Tim Lucas

          Hi Joan, the current working version is a reference design, I can’t divulge a price yet, eventually the cost to user will be negligible for wireless VR. In the short term it will come with a cost, for an accessory product compatible with current VR tech. Our core IP is a wireless compression algorithm, technically this could be inside any future VR compatible product and allow streaming capability, without after-market hardware.

          • Joan Villora Jofré

            Hi Tim, thanks for answering.
            As I have already explained to Daniel Fitzgerald, my interest was more than anything to clarify that the rumor of current versions to 1200 dollars should be a mistake, I am very interested in your product, in fact.
            I think it’s a great job and I hope you keep doing things just as good.
            Good work and good luck with it.

        • Daniel Fitzgerald

          Hi Joan, IMR CEO here. I noted the discussion on price below which some people have commented on incorrectly. We had discussed some B2B pricing options for a pre-release to certain companies for testing earlier in the year at CES. This is possibly where this higher price point has come out in some discussions but these numbers are incorrect for a product being released.

          See my next comment for details on our current status.
          thanks
          Daniel

          • Joan Villora Jofré

            Hello, Daniel, first of all, thank you very much for answering.
            I already imagined that it was some incorrect data, for that reason it speculated with ranges of price quite lower. As you have already seen, the problem is that the rumor has run that your link is by far the most expensive, which harms your product. If you allow me, I would advise that you deny it as soon as possible in future interviews. And if you could indicate a price range (approximate) that would be your link for current HMD, would help to silence this rumor.
            I’m following your work and I really like the results of your product. Good job.
            Thank you.

      • Peter Hansen

        Hi Tim, thanks for addressing my comments personally, that actually sounds pretty great! I have built a VR lab at a university with a tracking space of 16x8x4 m. Would you have any range concerns there? It is a single room setup.

        • Tim Lucas

          Hi peter, we have designed for the capability to do large room scale spaces, including the size you mention. There are a number of design considerations that I believe are unique to IMR, that allow for it!

          • Peter Hansen

            Thanks, Tim, can’t wait! :)

  • So their trick is hevc using an embedded encoder with just i-frames, but low quantitizer with a good bitrate?

  • Daniel Fitzgerald

    Hi guys. IMR CEO here. Firstly, a big thanks to Ben Lang who made the time to come out to our new Palo Alto office and review this. I wanted to briefly give a status of where we are currently at. We have generated a complete wireless reference design compatible for Oculus Rift and HTC Vive which we’ve demonstrated at a recent open night at IMR offices in Brisbane. As Tim mentions below we have achieved a fully wireless 4 player environment with player interaction. Link below to video of the event:
    http://www.immersiverobotics.com/blog/wireless-vr-tech-showcased-at-imr-open-night/

    We are demoing this is San Francisco next week and we will post some additional photos and video on our news page shortly.

    In terms of product release we are targeting 3rd party manufacturers and we expect an accessory product to be available later in the year at affordable pricing to customers like VR arcades.
    Please see this letter to the community I wrote from our new page below:
    http://www.immersiverobotics.com/blog/an-update-from-our-ceo-dr-daniel-fitzgerald/

    thanks
    Daniel