So Many H.264 Codecs, So Little Time
Though you would expect a CBR stream to be perfectly flat, in practice, they never are, and different encoding tools deliver different degrees of conformance to the target data rate. There are several points where MainConcept is much higher than Compressor (MainConcept peaks), but there are also several troughs where MainConcept is lower (MainConcept troughs). Had I run the comparison with Dicas, the results would have been similar.
How much did the difference in data rate affect the results? Theoretically, Apple’s quality should have been lower at the peaks and higher at the troughs, essentially balancing out any advantage that MainConcept/Dicas might have had. Not surprisingly, even though Dicas and MainConcept outperformed Apple in the most high-motion sequence, Apple’s quality did often exceed Dicas in low-motion sequences, although MainConcept won nearly every comparison. Anyone using the three encoding tools in the field would experience the same results, so they do represent reality, even if it is less of an apples-to-apples comparison than I would have liked.
In the highest-motion sequence in the test file, the Apple frame is noticeably more degraded than Dicas, which clearly trails MainConcept. While this isn’t totally obvious in the screen shots in the magazine, you should be able to view higher-resolution images on StreamingMedia.com in the web version of this article.
Next is a relatively low-motion clip with a hard-to-compress beige background from one of the data rate troughs in both the MainConcept and Dicas clips. As you can see, the Dicas clip shows a blurry face and vague blocks in the background, which were very noticeable during real-time playback, which cost Dicas points in the Motion Quality category.
The MainConcept frame retains a very high contrast with good quality and a relatively homogenous background, while the Apple clip is slightly blurry, a bit faded, and noisier in the background. Initially, I thought that the fading in the Apple clip resulted from the fact that I was using a test clip prepared in Windows and then encoded on a Mac. However, then I noticed that the Dicas clip, which was produced on the same Mac as the Compressor clip, was not faded. Then, I compared two clips produced in Squeeze (running on an HP Windows workstation), one using the Apple codec, the other the MainConcept codec.
The two clips produced with the Apple codec are both faded, despite being produced on different platforms, while the MainConcept clip, produced by Squeeze on the same computer as the Squeeze-Apple clip, retains very good contrast. If you’re using the Apple codec with any encoding tool, you need to watch for fading, and you may improve your results by boosting color and/or saturation either in your editor or in the encoding tool.
At the other end of the spectrum, if you’re used to bumping contrast and saturation prior to encoding and you plan on trying the MainConcept codec, run a few test encodes to make sure this is necessary. If anything, MainConcept—at least as applied by Carbon Coder—seemed to boost contrast during encoding, which generally produced a visually pleasing result.
What about motion quality? In general, the MainConcept clip exhibited the fewest motion artifacts, though there were two clips, both with very hard-to-compress backgrounds, that gave MainConcept fits: One was a seminar shot against highly detailed wallpaper, and the other was an action sequence shot against a bright red curtain.
At the other end of the spectrum was the Apple clip, which often exhibited artifact-related noise during real-time playback of high-motion clips. With hard-to-compress backgrounds, including the aforementioned seminar clip, the codec’s fading seemed to minimize some motion artifacts. In general, Dicas had across-the-board problems with clips with hard-to-compress backgrounds, especially the opera shot.
Overall, while MainConcept clearly delivers the best quality, the difference in most scenes is relatively minor, and viewers wouldn’t notice the difference without side-by-side comparisons, which, of course, they may never have. If you shoot with easily compressable backgrounds, limit motion, and are generous with your file bandwidth, all codecs should produce a good result from your viewer’s perspective. On the other hand, if you’re dealing with problem footage, or need to be stingy with your data rate, avoid Apple and prioritize MainConcept over Dicas.