David L
User
Feb 25, 2008, 9:42 AM
Post #19 of 30
(1824 views)
Shortcut
|
|
Re: [2ndMile] HDV WorkFlow - I need a little direction
[In reply to]
|
Can't Post
|
|
Just did a battery of tests, here is what I found. Bear in mind that the differences between these various ways are pretty small and the average person might not notice.
First I created a short, self-contained QT movie with some titles, rolling titles, panning shots, and stationary shots, then dropped that into Compressor 2 (I'm still on FCP 5.1.4). For all tests, I picked the 16:9 90-minute SD DVD preset, set the bit rate at 6.8 average, 8.0 max 2-pass VBR
First test I wanted to test the various algorithms within the deinterlace filter. I left the frame controls at their default, which is set at custom, resize=better, output=same as source, deinterlace=fast, adaptive details on, rate conversion=fast. Then tried each one of the four algorithms: blur, even, odd, sharp. I found the blur was the best. The odd and even algorithms resulted in some flickering (or shuddering or stuttering) on the rolling titles and pan shots, and I could see some jaggies in the stationary shots and titles. The sharp setting resulted in slightly more distinct titles but also resulted in mild jaggies on edges, such as the groom's collar. With blur setting, I couldn't see jaggies and the titles were reasonably distinct. So blur wins.
I then tried turning the frame controls off, with the deinterlace-blur filter on. That clip resulted in more flicker on rolling titles and more edge noise -- not really jaggies, just noisier. So I prefered the picture with the frame controls set to the default as noted above.
Then I tried frame controls set at automatic, with deinterlace-blur filter. I thought this picture was slightly better than the one with the frame controls at the default position.
Then I compared a clip sent straight from the timeline to compressor, versus a self-contained qt file droped into compressor. Both used the deinterlace blur setting and the frame controls custom as noted above. I couldn't see any difference. So I'm sticking to the QT self-contained route.
Next I tried a clip with frame controls set at automatic without deinterlace filter. Mild jaggies resulted. The deinterlace-blur filter seems essential for smoothing out the jaggies and flicker.
Then I read in the Compressor manual that it is better to use the frame controls deinterlace process than the deinterlace filter; and in order to do that, you have to set the frame controls output to "progressive", or it won't activate the frame controls deinterlace process. So I did that -- set the frame controls to custom, everything else default except output which I set to progressive, and did not apply the deinterlace-blur filter. This resulted in the sharpest picture yet (average person might not notice, but I did) -- however, it also introduced a mild shudder on panning shots and rolling titles.
So I added a deinterlace-blur filter, and that smoothed out the shudders for the most part, and made it ever so slightly less distinct, but resulted in the best picture overall, I think.
I also tried a test with the adaptive details off, but that resulted in too much motion stutter.
I also tried setting the output to progressive and then setting the frame controls deinterlace to Better -- after one hour, it was only done with 20% of a 1.5 minute clip, so I stopped it. Way too long. So "Fast" is the only feasible option there, I think.
So I think setting the frame controls to custom, output to progressive, deinterlace fast, then adding a deinterlace blur filter resulted in the best picture, with frame controls on auto and deinterlace blur filter a close second.
Then after reading Scott's post above, I tried nesting the sequence, adding FCP's deinterlace filter and broadcast safe, then output to qt file and did a comparison test ... I thought this resulted in either mild jaggies, or if compressor's deinterlace filter-blur was added to smooth those out, the picture became too soft. I may try the Nattress filter, though, perhaps that will do a better job than FCP's deinterlace filter. (another thought, would the broadcast safe filter have softened the picture? I doubt it).
So my conclusion: export timeling to qt movie self-contained, drop in compressor, 16:9 best preset, set frame controls to custom, output progressive, add deinterlace-blur filter ... results in the best trade off between picture sharpness and smooth motion. If sharpness is your main goal, then just going with the progressive route, and not adding the deinterlace filter, may be the best way.
(This post was edited by David L on Feb 25, 2008, 9:45 AM)
|
1
