X-bit labs

Ready to HD: Performance of Contemporary Graphics Accelerators during Video Playback (page 10)

Category: Video

by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov

[ 02/26/2007 | 05:24 PM ]

Video Processors and Image Quality: HQV Tests

Besides checking out performance of AMD Avivo and Nvidia PureVideo HD video engines, we want to see how they affect the image quality. To do this, we will utilize the popular HQV test (www.hqv.com) developed by Silicon Optix. It is intended for a comparative quality analysis of DVD playback. This test is an ordinary DVD with a handy menu system and onscreen tips, so it can be used not only on the PC, but in any device that can play DVDs.

The downside of this universality is in the certain bias of the end result because the user has to evaluate the quality of each feature visually, basing on reference images and assigning points. Fortunately, the test comes with detailed documentation that contains all the reference images, and if the tester is attentive enough, the influence of the subjective factor is rather low, and the results can indeed be used for making comparisons.

Tests from the HQV suite can be divided into several sections, each of which helps check out a certain feature of the video processor:

• The Color Bar/Vertical Detail , Jaggies Pattern 1 , Jaggies Pattern 2 and Waving Flag tests help evaluate the implementation of deinterlacing algorithms while watching interlaced video on a display device with progressive scan, like an ordinary DVD on a PC monitor or an LCD TV-set. The first test also serves to evaluate color balance, saturation, and overall image sharpness.
• The Detail Enhancement test helps evaluate detail enhancement technologies.
• The Noise Reduction and Motion Adaptive Noise Reduction tests check out the video processor’s noise reduction capabilities in still scenes and in scenes with motion, respectively.
• The 3:2 Detection test is for checking the video processor’s capability to convert film-mode video content (with 24 frames per second and progressive scan) into NTSC format (30fps, interlaced) if the content doesn’t contain necessary flags. This process is called 3:2 pulldown , and the recording of 24p content on a DVD with special flags that instruct the player’s video-processor to distribute frames to fields in the necessary order is referred to as soft telecine . If an MPEG-2 stream lacks such flags and the video processor is not smart enough to recognize the need to perform 3:2 pulldown, the picture will lack detail and will have visual artifacts (aliasing and moiré).
• The Film Cadence tests are a variation of the previous test, but with different pulldown sequences for different video formats, including 12fps for animation and 8fps for anime. As opposed to the previous test, there are necessary flags in the video, so it is the video processor’s ability to convert formats rather than to detect them that is tested here.
• The Mixed 3:2 Film with Added Video Titles tests allow to check out how well the video processor can perform “intelligent” deinterlacing. Sometimes subtitles or scene transition effects in 30i mode are put over an original 24p video. In this case the processor has to process different parts of the image in different ways, on a per-pixel basis.

Each test comes with a description and a few reference images each costing a certain amount of points. The tester compares these images with those he sees on the screen and awards points. The final result is the sum of all the HQV tests. The maximum possible number of points is 130.

Here are the results we’ve got:

As you can see, there is a very small difference between Avivo and PureVideo HD. Nvidia’s video-processor has a somewhat lower deinterlacing quality, especially in the Waving Flag test, but is superior to the ATI processor when it comes to detail enhancement algorithms. Although Avivo produced a sharper-looking picture in that test than the PowerDVD software decoder it didn’t allow to adjust that parameter and thus scored less points than PureVideo HD. On the other hand, the AMD solution was better at detecting the necessity to use 3:2 pulldown. Both video processors identified it with a delay of about half a second, but Avivo remembered the necessary mode and when the content was played again, it didn’t show moiré or aliasing in the first moments. That’s why it scored more points in this test. In the other tests the video processors from AMD and Nvidia deservedly scored the maximum amount of points possible. They produced a much better picture than the software PowerDVD decoder did even with enabled CLEV-2.

Unfortunately, the Chrome S27 graphics card didn’t pass the test. It produced a jittering image, probably due to the NTSC version of the HQV test disc. Having a video processor with good capabilities, the Chrome S27 gets nil, obviously due to its faulty drivers, which are a long-time problem of S3 Graphics. Although the Chrome S27 has good technical characteristic for an entry-level product, it cannot be recommended for playing video.