Test pattern exposes more issues with dejudder processing, but does it matter?
The HDGuru has posted video of a new test pattern that reveals additional video processing artifacts created by dejudder processing in 120Hz and 240Hz LCDs.
We've often complained about the video-like look of judder from 24-frame, film-based sources, producing a look some viewers prefer. For the record, we strongly prefer to leave these modes turned off.circuits like Samsung's Auto Motion Plus (AMP), Sony's MotionFlow, and LG's TruMotion found on those and other companies' equipped LCDs. Using a process called Motion Estimation/Motion Compensation (ME/MC), they remove some or all of the
But in addition to that videolike smoothing effect, the processing also causes further image degradation. We've documented numerous such instances, which often appear as halos, trails and other unnatural effects clearly visible in program material, especially during medium to fast movement, such as an actor turning his head quickly during a closeup.
Now a post at HDguru.com by reviewer and industry observer Gary Merson exposes additional artifacts caused by the processing. The artifacts are visible in a video of five LCD TV makers' dejudder-equipped LCD TVs (a sixth plasma TV's wedge lacks the flashing and much of the moire). They appear as unnatural flashes and tears in addition to extensive moire that looks like confused, curving lines. The test pattern in the video originated from a Blu-ray test disc by Spears and Munsil, a copy of which is included with the
Using a few of the 120Hz and 240Hz HDTVs I have in my lab at the moment, namely the the Samsung
The artifacts were entirely absent from a pair of plasmas I had on-hand, the
But does it matter? As usual with test patterns, the most important question is how well their results translate to real world video. The Guru's article says "This translates to a loss of fine detail such as film grain and overall image alteration." I've never noticed such a loss when comparing video processing circuits, but that could be because I'm not looking hard enough.
To find out, I stared hard at a relatively grainy Blu-ray, "Silence of the Lambs," and compared the TVs in my lab with dejudder turned on and then turned off. Frankly, it was nearly impossible for me to tell the difference. In all of the sets I looked at, any given TV with dejudder engaged looked just as grainy and detailed as the same TV with the mode turned off. Perhaps a more stringent test, with more source material and especially larger screens, would yield more differences, but either way I feel safe pronouncing the effects of those flashing dejudder artifacts as "subtle" at best. The same can't be said for the halos, trails and other dejudder artifacts, not to mention the smoothing effect itself.
As the Guru rightly points out, one of the benefits of 120Hz and 240Hz HDTVs is improved. Its effects are also quite subtle to our eyes--and those of --but assuming viewers want the extra resolution regardless of whether they can actually see it, we laud makers like Samsung, Sharp and Toshiba. Their 120Hz and 240Hz LCDs enable the viewer to disable dejudder and its attendant artifacts while keeping full motion resolution. Other makers, including Sony, LG and Vizio, cannot achieve full motion resolution without dejudder.
CNET's HDTV reviews don't currently incorporate tests from the Spears and Munsil disc, but I have seen these kinds of flashing artifacts before on other test patterns I use currently, such as the Film Resolution Loss Test from the HQV Benchmark Blu-ray and the moving monoscope from the FPD Benchmark Blu-ray used in our motion resolution test (for which the HDguru also deserves credit). I am considering revisions to CNET's HDTV testing procedure for 2010, however, and patterns from that Blu-ray could definitely make the cut.
So I'll ask the same question I asked of motion resolution: do you think tests like this are worthwhile to include as part of every TV review? Or are they just too esoteric, out-of-proportion and potentially confusing to be worthwhile? Let me know in comments.