De-interlacing in Apple Compressor
In our last segment, we identified the various de-interlacing options available in Apple Final Cut Studio, discussed when and why de-interlacing is an issue, and looked at de-interlacing in Final Cut Pro and when exporting via the Using QuickTime Conversion option. This time out, we'll look at the de-interlacing options available in Compressor.
You can get to Compressor by exporting a QuickTime Reference Movie from Final Cut Pro, and importing that into Compressor, and by using the Send To Compressor option in Final Cut Pro. In both instances, the sequence setting that you use will determine whether you can de-interlace in Compressor. Basically, if you're working with a progressive sequence setting in Final Cut Pro, you can't de-interlace in Compressor; if you're working with an interlaced setting, you can. Let's work through a scenario to see why.
Suppose you're producing with a progressive sequence setting and insert interlaced footage into the project. If you create a QuickTime Reference Movie to insert into Compressor, you're essentially saving the timeline into a file configured to the parameters of the sequence setting, which means a progressive file. If you use the Send To option, Final Cut hands off a series of frames configured to the parameters of the sequence setting.
This is shown in Figure 1, where I used the Send To command to send a progressive sequence (DV Progressive) to Compressor. When I click the project in the Batch Window and check the Inspector Pane, it's apparent that Compressor sees the video as progressive, matching the sequence setting.
The corollary is this. If you're working with a progressive setting, you have to de-interlace before you send the video to Compressor, because Compressor can't de-interlace progressive frames. If working with an interlaced setting, you can de-interlace in Final Cut Pro and/or Compressor, but you'll get the best results if you de-interlace in Compressor.
That decided, how should you de-interlace in Compressor: the De-interlace filter in the Filters pane, the de-interlace controls in the Frame Controls pane, or both? Let's check the Compressor help file, which says:
"For de-interlacing, the Frame Controls feature is recommended over this de-interlacing filter (a legacy filter) as it will always provide much higher quality."
That's easy enough. But which option should you choose: Fast, Better, or Best? Let's check back with the Help file, which says:
- "Fast (Line averaging): This option averages adjacent lines in a frame (note: essentially the same technique as Final Cut Pro).
- Better (Motion adaptive): This option offers good-quality de-interlacing for areas of the image that are in motion.
- Best (Motion compensated): This option offers higher-quality de-interlacing for areas of the image that are in motion."
And also: "Important: Using all Best settings may result in unexpectedly long processing times. If you are reducing the frame size in addition to de-interlacing the frame, Fast or Better will likely provide sufficiently high quality, depending on the amount of downward resizing."
This gets back to what I mentioned in the previous segment: The worst case for producing obvious de-interlacing artifactgs is when you present the frames at their original horizontal resolution, which is the dynamic used in all the tests files that I show. If you're scaling DV source footage down to 320x240, or 1080i footage down to 848x480 or 640x360, you should definitely de-interlace, but you'll see much less difference between the various alternatives.
OK, Fast, Better, or Best? Before looking at the qualitative results, let's gauge comparative processing time. Table 1 shows how long it took to render my 1-minute DV test file to 640x480 H.264 format using all three alternatives. As the table shows, times are for minutes and seconds.
So, the difference between Fast and Better is irrelevant, while Best is glacial. Let's look at some comparative frames to see how much benefit the additional processing time delivered. In our first comparison in Figure 2, the striped shirt presented a significant challenge for Compressor's de-interlacing capabilities, and Best certainly wins out by a noticeable margin.
Figure 3 shows a classic jaggy magnet, a golf club. Again, Best is noticeably better than Better and Fast; given the small difference in processing time, Fast doesn't seem like a reasonable option. Two golf shots in a row. Raise your hand out there if you think I can write off my next set of golf clubs for tax purposes.
Figure 4 is another look at the guitar shot that I showed back in the first segment, two weeks ago. Again, Best looks superior to both other options.
Figure 5 is our skater. Here, Better is nearly indistinguishable from Best, and clearly better than Fast.
Overall, with a processing time about 17 times longer than Better, Best is aptly named when time isn't of the essence and your footage contains jaggie magnets like guitar strings, golf clubs, skateboards, and the like. For most footage, however, Better likely provides sufficient quality and is much more efficient. Significantly, all of the Compressor presets that I checked use Better, not Best, so that's where I would start.
I wanted to circle back to a point that I made last time. If you need to integrate interlaced footage into a progressive project, your highest-quality option is to convert the footage to progressive first, using an interlaced sequence setting in Final Cut Pro and de-interlacing in Compressor. You can see this in Figure 6, which compares this technique with the results achieved by adding the interlaced file to a progressive timeline and applying Final Cut Pro's De-Interlacing feature.
As you can see, the Compressor - Better frame on the left is clearly more vivid and clear, and given the similarity in processing time, the superior option.
That's it, I'm going shopping for those golf clubs. If you'd like to view some of the files that I created performing this research, check out this link. The files are very large (5Mbps) to preserve the quality differences, so allow some time for the files to download.