ArcSoft Claims HD From SD With Nvidia CUDA

A new software suite from ArcSoft promises to turn standard definition into high definition.

We hear all the time about the upscaling quality of high-definition players and how they handle the conversion of a 480p image to 1080p. It’s one measured areas that a home theater aficionado (at least those who still have sizable DVD collections) pays attention to. With that in mind, we raise eyebrows at ArcSoft’s claim that its new software can scale DVDs to near-HD levels.

ArcSoft promises that its SimHD plug-in for ArcSoft TotalMedia Theatre multimedia player will work upscaling magic by utilizing the Nvidia CUDA parallel computing architecture to “solve complex calculations in a fraction of the time required on a CPU.”

TotalMedia Theatre takes advantage of the GPUs from the GeForce 8 Series onwards for SimHD‘s “intensive post processing algorithms.”

"Our newly released upscaling technology, ArcSoft SimHD, is available now in retail to allow viewers to obtain an HD-like viewing experience on the PC from the existing standard DVDs,” said George Tang, ArcSoft VP and GM.

“What a great way to upgrade your existing library of DVDs! All you need is TotalMedia Theatre and an Nvidia GeForce GPU, and you can instantly turn your movies into near-HD quality,” said Michael Steele, GM of visual consumer solutions at Nvidia.

Although filters and post-processing do help an image, we’re extremely skeptical about any claims of turning a standard definition picture into one that’s even near the fidelity of high-definition.

A 1080p Blu-ray Disc is six times the resolution from a DVD, so how SimHD is able to make up for that difference -- at least what is perceptible to the eye -- is beyond us. The example image included by Nvidia and ArcSoft shows image tweaks, but not a shift in resolution (which is to be expected).

TotalMedia Theatre with SimHD technology is currently available as a TotalMedia Theatre plug-in for end users, and we’ll have to check it out for ourselves.

Marcus Yam
Marcus Yam served as Tom's Hardware News Director during 2008-2014. He entered tech media in the late 90s and fondly remembers the days when an overclocked Celeron 300A and Voodoo2 SLI comprised a gaming rig with the ultimate street cred.
  • makotech222
    And what will be the size of the new DVD file? Still fits on a regular DVD? I hope so...
  • ohim
    Nomatter what technique you use you can`t turn SD material into HD ... SD PAL material has 720x576 pixels while 1080i has 1920×1080 no matter what filters you use you can`t "invent" out of nothing more than double the missing pixels.
  • eklipz330
  • KyleSTL
    It's real, it just has unrealistic claims about enhancement. 720x480p will never become true 1920x1080p no matter how you slice it, but I'm sure the image quality certainly won't suffer from the post processing. Same with stand alone DVD components and receivers with the ability to upscale. PowerDVD has a similar feature called TruTheatre HD, the elimination of the 'e' at the end of words is becoming the new 'Extreme' or 'Turbo' or 'i'. It just makes it that much cooler.
  • Actually, there are lot of hidden information between pixels which CAN be used to extract resolution out of any given image. Every pixel contains information about every other pixel near it, and combined with decompressing of codecs the result can reveal/"invent" a lot more detail than the original picture actually has. It's sort of like filling in boxes in sudoku puzzles using logic and fuzzy guessing. This technique is different from up-scaling and it is called super-resolution. There is a lot of research going on in this field right now. That said, I don't know how effective Arcsoft's software is.
  • BigBag
    Shouldn't we be able to turn HD content into super HD using a similar process then?
  • celebrity
    My question exactly. But I think it depends on the codec involved and how much extra information is stored within the codec. I think that the newer codecs used for HD is much more compressed and less likely to be able to extrapolate the missing information. After all, why can you get almost DVD quality out of a file at times less than 1/8 the size when using DivX (H264, which is HD broadcast)
  • ooo
    Well i wonder if they are using the technology behind NeuView Media player AKA Pixel Fusion Technology cuz with this little and free application you can change any 640x480 to looks like 720p sources applying the effect at maximum from the menu... and sources HD movies like 720p KingKong looks super should try this1 too...
  • It might not entirely be impossible, however when applying the procedure it might be that the quality gained is unlike the original.
    First there's linear and cubic interpolation;or perhaps quarter pixel data that can be extruded to a cluster of 4 pixels.

    A more complex,and CPU intensive procedure might include extracting 720x480 to 1440x960. A cluster of 4 pixels will turn into 16 pixels.
    Depending on the shape of the object, the GPU calculates how an object which crossed the 4 pixels will display on 16 pixels.
    If it's crossing in a straight line, (or in a curved line) it will look much like enabling AA 2x in games. The extraction of 4 to 16 pixels might be more than simple upscaling happening all over the screen.

    A third way this technology could work, is if the images processed are compressed with a lossy encoder like divx or xvid.
    Both codecs can successfully identify cut foreground objects, sometimes store them as high quality Jpegs, and slide them across the screen,to eg show an animation of an object sliding across the screen.
    The data they gather of these objects sometimes surpasses the quality or resolution of the display they are displayed at (eg as they move across the screen they are also scaled to smaller objects).
    Instead of recalculating an object every frame,DivX and XviD can 'scale' that object larger or smaller,and keep the highest detailed version in the memory using that to display even at lower resolutions.
    If the software is capable of calling these objects, it could load the data it has about this object into the memory, and perhaps display it in a slightly higher resolution as the original display resolution
    (if you get what I mean).

    Latter 2 examples are very uncommon and probably extremely high CPU/GPU intensive tasks. So I think this methode is nothing more than extracting quarter pixel data to 2x2 pixelarray, and using the simple AA 2x approach.
    The original might look different from the upscaled version, especially if the image contains lots of data & detail!
  • gimarbazat