Summary: Think upsampled DVD is just as good as Blu-ray and HD-DVD? Heard amazing things about the PlayStation 3? Interested in comparing Avivo and PureVideo? We've got all of that in this Autumn's Video Processing Face Off!
Think upsampled DVD is just as good as Blu-ray and HD DVD? Heard amazing things about the PlayStation 3? Interested in comparing Avivo and PureVideo? We’ve got all of that in this Autumn’s Video Processing Face Off!
Autumn is on its way. The sun is setting earlier, things are starting to cool down, and before you know it, the leaves will start to change color and we’ll be in winter. With all of the Summer movie blockbusters finally making their way to home video, and an upcoming surge of reunions and parties with the usual holiday get-togethers (i.e. Thanksgiving, New Year’s, Christmas, Hanukkah, Kwanzaa, Bodhi Day, El ul-Fitr, Boxing Day, Festivus, Winter Solstice, Chrismukkah, or whatever seasonal holiday I haven’t mentioned that you celebrate), there’s no better time to get your home theater up and running in tip-top shape. What better way to spend a brisk autumn or winter night than having some friends over for dinner and a movie?
Of course, being a FiringSquad reader, you’re not satisfied with just “another” article that runs a bunch of tests from the HQV Benchmark DVD/Blu-ray/HD DVD. The HQV benchmark is one of the best tools in the industry for evaluating video performance, but that doesn’t tell you everything.
Likewise, as a FiringSquad reader you won’t be impressed by a reviewer claiming that “upsampled DVD is nearly as good as Blu-ray and HD DVD,” when they’re testing with a 42-inch 720p Plasma, or talk about “normal viewing distances” without even mentioning SMPTE or THX horizontal viewing angles, or simply have poor eyesight.
Just think about the last time you read someone claim that there’s no 1080p content other than Blu-ray or HD DVD. As a FiringSquad reader, you should know that the vast majority of 1080i broadcast television contains a full 1920x1080 pixels for each frame of original frame that can be reconstructed into full 1080p60 using only the data that’s available.
Without further delay, I present to you, FiringSquad’s Autumn 2007 Video Procesing Face Off.
The benchmark for video processing has historically been “Hollywood Quality Video” from Silicon Optix. These processors are found across a wide range of products ranging from the $500 Toshiba HD-XA2 (ReonVX) as well as in dedicated video processors starting at $5000 (Realta). To test the upscaling and detail enhancement performance of the Realta, we used the PlayStation 3 as our source with all sharpening and noise reduction disabled.
Finally, we included the Sony PlayStation 3. Although the Xbox 360 offers a wider selection of games than the PS3, Sony’s console is in a completely different class when it comes to home theatre and hi-fi performance. Like many premium audiophile CD players, the PS3 is able to resample conventional music CDs to 88.2 kHz (optical out) and even 176.4 kHz (via HDMI). As a Blu-ray player, the PS3 is one of the fastest players on the market in terms of BD-Java performance and compatibility, and even features high-end capabilities such as 1080p24 output. Although the PS3 can convert 1080p24 recorded films (i.e. Hollywood movies) to 1080p60, it cannot deinterlace 1080i60 to 1080p60 at this time. Sony engineers have stated that they are assessing 1080i60 to 1080p60 deinterlacing internally. Since 1080i to 1080p upconversion is compute intensive, Sony needs to make sure that there is enough hardware resources available to accommodate updates to the Blu-ray specification. At the moment, however, Sony does use the Cell processor to provide advanced DVD upscaling beyond the traditional bilinear filtering.
Although we had planned to include the Toshiba HD-XA2 (HQV ReonVX processor and HD DVD player) and the Microsoft Xbox 360-HDMI with the HD DVD upgrade in this comparison, neither company was able to provide us with a sample in time for participation. We hope to revisit Toshiba’s HD DVD players and the Xbox360 in a future article. By that time, we hope to include a comparison between several dual-format players.
Sony PlayStation 3 (Firmware 1.9)
To test the challenge that upsampled DVD is “nearly as good” as Blu-ray, we decided to use Pirates of the Caribbean: Dead Man’s Chest. When it comes to picture quality, not only is this one of the best Blu-ray discs on the market, it is also one of the best DVDs on the market.
This comparison shows how far technology has gotten. Although the AMD Radeon HD 2600 offers the blurriest picture in this group, it’s important to keep in mind that the Radeon HD 2600 is already superior than the typical video processor found in most HDTVs on the market. Unfortunately, the Radeon has a fixed detail enhancement algorithm. With 51% sharpening (the half-way point), NVIDIA PureVideo provides a very good image. Even then, the PlayStation 3 seems to outperform NVIDIA’s upsampling quality by just a hair. This is a true testament to the power of the Cell architecture and how easy Sony has made it for the end user.
While we always value user-adjustable sharpening (as NVIDIA and HQV Processors allow), Sony’s default settings provide very good picture quality with no glaring artifacts.
For this particular scene, the PS3’s 1080p upsampling is actually a hair sharper than the HQV. In fact, you just need to look at the Kiera Knightley’s hair and eyelashes. The PS3’s output does produce a bit of aliasing (most easily seen in the strand of hair crossing in front of the eye) – the HQV provides a smoother image. The Realta’s detail enhancement that provides added local contrast giving a “punchier” overall image.
But you cannot make gold out of lead. There’s simply no comparison to the same frame of film from the Blu-ray edition:
Our next shot is a well-lit outdoor close-up. Continuing with the theme, we again see ATI offering an image with no jaggies, but also a slight blurrines in comparison to NVIDIA, Sony, and HQV processing, which offers more conservative image enhancement without the artifacts.
The PlayStation 3 continues to offer a strong showing, and again we feel that the PS3 slightly outperforms the HQV and PureVideo solutions when it comes to sharpness. Pay close attention to Johnny Depp’s eyes. Once again though, upsampled DVD is still no match for a native Blu-ray disc.
If you look at the third triangle in the lower right corner, you can see that the PS3 is the sharpest of the bunch when it comes to DVD upsampling.
In the last two shots, we focused on two actor close-ups. The advantage of high-definition is more prominent in wide-angle shots such as this one.
Once again, it’s hard not to be impressed with the PlayStation 3. This definitely was not what we had expected to see going into the round-up. One thing to notice with the PS3 is that on the rear end of the ship, there is a slight sharpening halo which shows up as a bright border. This is a sign of over-aggressive sharpening on the PS3. To repeat myself a third time, the Blu-ray edition is four times sharper than the DVD release. In the Blu-ray version, you can even see the individual rungs on the ladder on the side of the ship!
DVD Upsampling Summary
The next time someone claims that upsampled DVD looks just as good as Blu-ray, you should point them to this article. Sure, you can do a lot of hand waving or argue that if you sit 25 feet away from your 25” TV that DVD and Blu-ray are identical, but for rest of us, high-definition media such as Blu-ray and HD DVD will clearly provide a step-up in terms of performance.
We were impressed with the PS3’s high-quality upsampling for DVDs, but for Blu-ray playback, the PS3 doesn’t have the same level of post-processing capabilities. In fact, the PS3 has no Blu-ray image enhancing technologies. When fed a 1080p24 source, the PS3 can output 1080p24 or 1080p60, but it can only output 1080i60 content at 1080i or 720p.
We focused our testing on three key capabilities: noise reduction, diagonal filtering, and the quality of 1080i60 deinterlacing including HD inverse telecine. Fortunately, AMD AVIVO, NVIDIA PureVideo, and Silicon Optix HQV technologies all feature HD inverse telecine.
For unclear reasons, by default, the AMD Radeon will not support full 1080p playback of Blu-ray discs with PowerDVD. Although the display will sync appropriately at 1080p (including all of your applications and DVD playback), with Blu-ray sources such as the HQV Benchmark Blu-ray Disc, there will be black borders all the way around. This does not seem to be a problem with Windows Media Player.
Searching the forums, we were able to find some discussion about a “VForceMaxResSize” registry tweak to enable support for full 1080p. On our setup, we did not have any luck with this setting. As a result, we were unable to test the Radeon HD 2600 in full 1080p.
Although there has been some recent concern about “cheating” with drivers, the real issue is that noise reduction continues to be a difficult algorithmic problem. A professional level noise reduction algorithm such as that found in an Algolith Mosquito can cost as much as $3000 for a device. In the consumer area, HQV tends to offer the best noise reduction capabilities thanks to its true per-pixel processing.
One of the best overviews of noise reduction can be found at http://www.hqv.com/technology/index1/noise_reduction
It comes from HQV.com, but it is very accurate in describing the issues at hand. I know, because I did the storyboard for that video. In general, the more noise reduction you apply, the more detail is lost.
NVIDIA PureVideo uses an adaptive temporal filter. This works by averaging pixels over several frames. When the noise reduction is turned all the way up, this produces some smearing of the image (which can be seen in the images that were distributed by AMD). At lower noise filtration, this effect is diminished, and by default NVIDIA does not enable noise reduction. ATI Avivo does not allow end-users to adjust the level of noise reduction.
The difference between AMD, NVIDIA, and HQV’s noise reduction techniques is how well the algorithms are at determining motion. Adaptive spatial-temporal noise filter works by identifying the pixels that are in motion -- these should be exempt from temporal noise reduction. The remaining pictures can then be filtered. HQV detects motion on the per-pixel level, allowing it excellent discrimination between motion and static elements of an image. NVIDIA detects motion on a region-based level which allows it to distinguish between motion and static elements of an image, but with less precision. AMD’s solution also features a spatial-temporal noise reduction algorithm, but it does not allow you to adjust the intensity of the effect.
The Realta HQV is the king of noise reduction in this round-up with NVIDIA coming in second. AMD appears to offer similar quality to NVIDIA, however the Radeon is displaying an image with less than full 1080p resolution. Any time an image is downscaled, noise gets filtered out – this prevents fair comparisons from being made as we are now comparing kumquats and oranges.
In this test clip, there is a split screen with a static right side of the screen. This is intended to provide a reference image. As a static image, very little noise filtration should occur. In general, filtering noise from static images results in a loss of picture detail unless you are dealing with sophisticated wavelet-based algorithms such as that found in the digital photography world (Noise Ninja, Neat Image, etc.)
Once again, direct comparison with AMD images are impossible. Since the image is downscaled, an equal sized crop will provide the appearance of a sharper image. NVIDIA’s noise reduction (when set to 51%) preserves most of the detail in the image. It does appear to be a smidge blurrier than the still image. The HQV noise reduction algorithm set in its default “medium” intensity provides a gentle reduction in noise, even in these static areas, without any significant loss of resolution. When set to “high” intensity, some details are beginning to disappear. Even so, it’s still sharper than NVIDIA’s solution (look at the deep purple veins in the lower right petal) while offering a creamy, smooth reduction of noise.
There is no question that Silicon Optix’s HQV noise reduction remains one of the best in the industry. We hope to take a closer look at products such as the Algolith Mosquito or Flea HDMI in the future.
A good overview of deinterlacing can be found in the video clip at: http://www.hqv.com/technology/index1/deinterlacing.cfm
We recommend viewing that link before continuing.
In general, interlaced video only contains half of the data you need to create a full image. The goal of deinterlacing is to preserve all of the detail that is available in the original image and then mathematically estimating the data in the missing areas. The better the video processor is at estimating the missing data, the better. On the other hand, a video processor should be smart enough to avoid estimating data when the real data can be recovered in the original signal. This requires identifying objects that are in motion and those that are not. This means that a video processor should have to discard data from moving pixels.
Motion Detection Torture Test
The HQV Benchmark Blu-ray features a classic video processing test. You have a high resolution SMTPE test pattern and a spinning white bar. A good deinterlacer will deinterlace the moving bar while leaving the background intact. Since the majority of video processors on the consumer market today discard as much as half of the resolution, HQV’s test clip is designed to showcase just how poor some video processors can be.
In the corners and center of the screen there are areas that contain fine detail (single pixel high stripes) which are nowhere close to areas of motion. On a poor quality de-interlacer, these areas will either flash or be blurred to gray.
Using this test, HQV and PureVideo processors score perfect scores. In order to challenge the video processors more, we use a tougher deinterlacing test. Instead of having the spinning bar on only one area of the screen, we have a larger one in the middle that crosses the central portion of the screen. To date, only HQV processing and Sharp’s CV-IC System III (which due to cost reasons is now only available in the Aquos LC-65D90U and XV-Z20000 1080p projectors) have successfully passed this torture test. Using the default “automatic” mode of the ATI produces flashing boxes (bad), but users who switch to the “Vector Adaptive” mode will find that the correct result is seen.
Diagonal filtering refers to a video processor’s ability to prevent jaggies
Although it is again, difficult to compare the AMD AVIVO results against PureVideo and HQV due to ATI’s mandatory resizing with PowerDVD, all three manufacturers provide virtually identical results. NVIDIA appears to have the advantage over HQV processing by just a hair, however if you use a more challenging test clip:
You can see that NVIDIA has a slightly blurrier image (one method used to reduce the appearance of jaggies). The edge goes to HQV.
Finally, in this round-up, we’ll be taking a look at the H.264 and VC-1 Compute Offload of AMD AVIVO and NVIDIA PureVideo in the video flagship HD2600 and 8600GT. MPEG-4 AVC, also known as H.264, represents one of the most computationally intensive consumer applications today. Even fast Core 2 Duo or AMD Athlon64 CPUs can struggle under heavy load. In response to this, both NVIDIA and AMD have developed GPUs with full H.264 decoding. AMD has gone a step further with full VC-1 decode. VC-1 is a SMPTE-certified video CODEC originally developed by Microsoft. The majority of HD DVD and Blu-ray content is encoded in VC-1 and MPEG-4 AVC with the MPEG-2 also being used on a number of Blu-ray titles.
For our testing, we used an Intel Core 2 Duo E6600 with 4GB of RAM running Windows Vista Ultimate 32-bit. We used Cyberlink PowerDVD for our hardware acceleration evaluation (recommended by both NVIDIA and AMD). PowerDVD will automatically take advantage of any hardware acceleration it finds so we used WinDVD to test the software decoding performance. These results are not applicable to the Radeon HD2400/HD2900 or the GeForce 8500/8800.
As you can see, the computational demands for 1080p h.264 is significant even for a relatively fast CPU such as the Core 2 Duo. Even with an animated film such as Chicken Little, CPU utilization was as high as 56%! The “full hardware” decode really works for both NVIDIA and AMD processors. In fact, the remaining 5 to 6% is likely going toward the audio and AACS decoding of the Blu-ray discs.
AMD’s video processing capabilities with VC-1 are worthwhile to note. While NVIDIA only offers partial VC-1 acceleration, AMD offers a complete VC-1 offload. For premium desktops this may not be a critical feature (there’s plenty of CPU horsepower to go around), but for thermally constrained systems or laptops, AMD’s solution is the better choice. Unfortunately, they currently do not provide full 1080p display.
When it comes to DVD upsampling, it’s hard not to be impressed by the PlayStation 3. We think that it provides a superior output to NVIDIA and AMD’s technology and it even holds its own against the flagship HQV technology from Silicon Optix and Teranex. But that’s just for DVD playback.
Once you make the plunge toward HD DVD or Blu-ray, additional challenges come into the mix. We still believe that the PlayStation 3 represents an excellent stand-alone Blu-ray player given that the majority of content that most people watch comes from 1080p24 Hollywood films – things that the PlayStation 3 does well with. With PS3 in the $500 price range, about the same price as an a high-end HD DVD player, the console suddenly looks like a great bargain for anyone who can take advantage of upsampled DVD, Blu-ray playback, and an entire gaming console with exclusives like Gran Turismo 5, Final Fantasy XIII, and Metal Gear Solid 4. It’s funny how history repeats itself -- the reason to buy PlayStation 2 was Gran Turismo 3, Final Fantasy X, and Metal Gear Solid 2…
Noise reduction continues to be the domain of the videophile-elite and going with an HTPC solution from AMD, NVIDIA or a dedicated video processor that is powered by HQV Technology are a few options. The price premium of Realta HQV over its consumer competitors is well deserved for those looking for a robust noise reduction. When combined with Realta’s generally conservative but effective detail enhancement and second-to-none per-pixel deinterlacing, it’s clear why HQV’s experience as the video processor used by NBC, CBS, ABC, and CW and other real Hollywood production environments.
We continue to prefer that the noise reduction be done by the end-user rather than the studio at the time of mastering. Careful attention is required at the studio when mastering these HD DVDs and Blu-ray discs, but it makes the most sense to work on capturing all of the detail that is available in the master print while adding as little noise as possible. Noise that has already been introduced during the compositing and filming stage should be left on the recording. The optimal amount of noise reduction will vary on the user’s display equipment, and if studios begin to do additional noise reduction before encoding, we will run into the same problem we had with DVDs and “excessive edge enhancement” – it’s not excessive if you were using an SDTV with an S-Video connector.
Finally, when it comes to CPU-offload of computationally intensive video decoding, AMD’s UVD with full VC-1 and AVC decoding makes it a great choice for dedicated home theater PCs and notebooks. Unfortunately, AMD’s default driver settings result in less than ideal performance and an inability to display the full 1080p image. Until this is fixed with the default driver set, we can only recommend AMD’s solution for consumers running resolutions less than 1080p (where the black borders are not a problem).
In closing, even though the PlayStation 3 is a mediocre game machine at this time, its CD, DVD, and Blu-ray performance is stellar, making it a smart decision even at its current price. AMD’s solution holds the crown for video decoding performance, but driver idiosyncrasies prevent us from making a full recommendation. We do think that every extra bit of performance helps, particularly with a laptop. So when a mobile version of AMD AVIVO HD is available, we’d lean in that direction.
NVIDIA offers a well-rounded solution, losing to AMD in VC-1 decoding performance and 1080i deinterlacing (when the Radeon is configured for vector adaptive deinterlacing). AMD’s advantage with 1080i deinterlacing is important for 1080i content, but when it comes to Hollywood movies, NVIDIA’s sharpening and noise reduction capabilities make a bigger difference in picture quality. Realta HQV continues to show the strength of dedicated video processors, but at the moment, the Realta is still too expensive for most end-users. Syntax/Olevia has a LCD panel with Realta processing; we expect more to be announced in the future.
|© Copyright 2003 FS Media, Inc.|