Choosing a Video Codec

That wasn’t the real surprise, however. The shock of the study for me, having penned an article entitled "MPEG-4 is Dead" a few scant months earlier, was how much the MPEG-4 codecs have improved in their latest iterations. The same goes for Flash. While quality didn’t quite meet the hype from the vendors who support them, MPEG-4 and Flash codecs are close enough in quality not to rule them out if other factors (device playback, etc.) dictate their use.

It’s Gonna Cost You (You Can’t Get There from Here)
We said that Real performed very well in our trials; the relevant question is how well, and how much higher must you boost the data rate of video files you create using other codecs to match Real’s quality. We analyzed this for several key technologies; here we see an analysis of the (CBR-based) Flash 8 Video Encoder.

In Figure 1, the original Real file, at 300Kbps, is in the center of the top row, while the two-pass, variable bit rate (VBR) VP6 file is beneath it. Both are surrounded by Flash 8 files produced at 450, 500, 550, and 600Kbps. Though it’s probably tough to tell in this picture, the mosquito-like artifacts around the lettering in the 450Kbps clip (and not present at all in the Real clip), are only slightly less prominent in the 600Kbps clip. The bottom line is that at twice the data rate, you still don’t get the quality.

Figure 1

While the data rate cost for Real-equivalent quality is less in some low-motion sequences, data rate won’t help with some scenes; your only option is to redecorate. More on that later under the subheading "You Look Marvy Against Black."

Your Mileage May Vary
We saw great variability in results over the different categories in the study. For example, in the Flash study (which included Real Video for context), Real won the 500Kbps business video category by 39% over the VP6 codec produced by On2’s Flix Pro Encoder. However, in the next category, Action, the VP6 video bettered Real by more than 32%.

For some content providers, that may be enough to convince them to switch codecs for different projects; for others, it’s just a warning that they may have to boost data rate or change their shoot to produce sufficient quality. Everyone should understand that there is no one-size-fits-all codec that encompasses the complete range of video, animation, and pan-and-zoom clips.

Choose Your Encoder with Care
To produce optimal quality, you don’t just choose a codec; you choose a codec/encoder combination. This may come as a bit of a surprise, because most encoding tool vendors incorporate pre-built, self-sufficient chunks of code from the codec vendor to handle the encoding chore. The encoding tool inputs the source file and performs the scaling, deinterlacing, and other processing selected by the user, and then hands off the resulting file to the encoding module for compression.

At first glance, this dynamic would seem to leave little room for difference in encoding output quality, but this wasn’t the case in our tests for two reasons. First, scaling and deinterlacing quality are absolutely critical to the quality of the final encoded file, and this varies among the encoders (see section under the subheading "It’s All About Deinterlacing"). However, even when we input the same file, already scaled and deinterlaced, into multiple encoders, we saw very different results.

Though Figure 2 is a bit extreme, it kind of proves the point. One of the batch encoding utilities produced the day-glo image on the right, while the Microsoft encoder was much more accurate on the left.

Figure 2

In other tests, the encoder that performed best when encoding Windows Media files turned out to be the worst encoder for Flash VP6 files. Equally interesting was that the Real Encoder bested all three batch encoders in quality tests, though one was very close, while Microsoft’s Windows Media Encoder placed third out of four Windows Media encoding tools.

In other trials, the encoder that produced the best VBR VP6 files produced the worst constant bit rate (CBR) files for low-motion footage. Overall, choosing the right encoding tool is as important as choosing the right codec.

Get to Know Your Encoder
As streaming production becomes more prevalent, software vendors make their encoding tools more user-friendly. Of course, this doesn’t mean that they’re foolproof, and often missing one or two key buttons or checkboxes can make the difference between great and awful quality.

Take the image in Figure 3. On the left is relative beauty at 500Kbps, on the right, a rebuke from the boss. The difference? An obscure deinterlace setting hidden in a Filter screen that you don’t see when accessing the Flix Pro main encoding controls (Figure 4). This isn’t a problem with Sorenson Squeeze, which automatically enables deinterlacing, but there is another "gotcha" we talk about in the report that can wreak almost equal havoc on your video quality when producing Flash FLV files.

Figure 3

Figure 4