DV conversion to MPEG

[AndrewDV]

Hello,

I have some video I captured using DV. After doing some editing, I am now ready to send this to VCD and SVCD.

First, I used the Ligos engine from MSP6.5. That did not turn out bad, but a bit less than expected.

Next, I tried using TMPGEnc. The results were good. However, the visual quality was slightly less than what I am used to from using PIC MJPEG (Q=20).

I'm assuming the lower quality (not much) comes from the DV compression. Does anyone know any settings/tricks for getting better MPEGs from DV?

I'm wondering if I should just stick with PIC MJPEG. Any thoughts?

Thank you

[wkulecz]

Whats the data rate for PIC MJPEG Q=20? for 720x480 DV its ~3.6MB/sec ~= 13GB/hr. For HuffYUV lossless its ~ 10MB/sec for 704x480 or almost 40GB/hr. Only you can decide if the quality of the resulting MPEG is "worth it" for the extra resources required.

For me with TMPGEnc12d making VCD format MPEG1 the difference between DV captures of SVHS tape vs HuffYUV captures of the same SVHS tape was nowhere near "worth it". YMMV.

--wally.

[Phil Pugh]

Yep - if youv'e captured from a VCR tape it is quite likely that you have already lost a bit of resolution. The better the source tape and recorder/player the less information (e.g. resolution) you are going to loose. You can also add noise into the process - which upsets MPEG (1or 2) processing quite a bit.

[Dr Mordrid]

Some of miniDV's problem with MPEG encoding is endemic in its format. When encoding to MPEG its 4:1:1 colorspace has to be converted to 4:2:0. This alone can cause some artifacting, but there is more.

Because a 4:1:1 colorspace smears each color sample over 4 horizontal pixels there is less color information than would normally be present in the 4:2:2 YUV sources MPEG was originally set up for. This causes problems in MPEG encoding because these smeared color samples can then orgainize, in regions similar color or shade, into rather large and annoying block artifacts.

In TMPGEnc there are some settings that can help;

First is a setting on the Environment Settings/General tab for "Interpolate YUV data from 4:1:1 to 4:4:4". Select this when using DV sources.

Second is a setting in Settings/Quantizer Matrix. On the bottom of the page is a checkbox for "Soften block noise" and some parameters. Try the defaults first and see how it goes. Play as needed. What this does is blur the video slightly to make artifacts less obvious.

There is also a noise reduction filter in Settings/Advanced, but I'd save that as a last resort. It too reduces sharpness a bit in an attempt to make artifacts less visible.

While burring & such may go against the grain a bit sometimes it's the only way around DV's limitations without resorting to capturing the DV footage as analog (which again blurs it a bit) or shooting known problem scenes for DV with an analog cam, both of which I sometimes have to do.

None of these TMPGEnc features is a real solution. Only the miniDV codec evolving into a 4:2:2 color sampling rate can fix it totally. Even then very few encoders, even commercial ones, offer the help with DV that TMPGEnc does.

One thing I've found useful when editing problem DV footage is my RT-2000 editing board. While it's not cheap and I have to use Premiere (Yuck!!) the fact that it upsamples all DV footage internally to 4:4:4:4 RGBA (RGB video with an alpha channel) for editing/compositing makes adding effects much cleaner. One can then frameserve this 4:4:4:4 RGBA stream into TMPGEnc using the AVISynth plugin and scripting tool. The resulting MPEG's are VERY nice.

Dr. Mordrid

[AndrewDV]

Thanks for the info!

FYI, the source is very clean Hi8 footage. I've been converting it to DV. I've been playing around with different methods of doing this. So far, my favorite is the Hollywood DV Bridge. I'm hoping to get some time to share my opinions about it.

Wally,

PIC MJPEG is pretty easy for me to work in. With 130GB on a FT100, I can get about 3 hours of footage. Of course, that goes up quite a bit with DV.

What I am noticing with DV is that the quirkiness of analog capture goes away. This is a nice benefit. MSP6.5 also seems a bit more speedy as well.

[wkulecz]

My problems with the Dazzle DV Hollywood Bridge was gray blocks on output back to tape. I had no problems with captures in somewhat limited testing.

I had hoped to use it instead of tying up my camcorder while producing VHS tapes. Had to return it. I've seen claims that Dazzle has made a revision that fixes the problem but haven't had time to waste buying and possibly returning a third one.

Now that my initial experience with DVD-R suggests I'll still be making tapes for a lot longer than I'd hoped, I may give it a try again in a couple of weeks.

--wally.

[AndrewDV]

Hi,

I'm not sure how to tell if I have the newer model or not. I did read several articles saying exactly what you mentioned. It sounded like a software update was going to be made for current users (flashes the hardware). So far, everything is pretty happy.

[Dr Mordrid]

Just a postscript on artifact handling & prevention in TMPGEnc;

As I mentioned before some blurring can prevent block artifacts at sharp or high contrast edges. One other way to do this: the "Settings/Advanced/Sharpen edge" filter.

No, I'm not being contradictory. The TMPGEnc "Sharpen edge" filter can be used with negative numbered settings, which result in a very controllable blur being applied to the video. This blur is also multi-directional in that it has horizontal and vertical settings. VERY handy indeed.

Most times I try to keep it to no lower than -20 in the horizontal and -6 in the vertical, but you can adjust it to suit your eye and needs.

Make sure to check the "Field base" box at the bottom of the dialog if the video is interlaced.

Such a slight blur can do a lot of good with problem sources, DV etc. if you don't over do it.

Dr. Mordrid

[AndrewDV]

Hi,

Just an update on this. I tried the options pointed out to me here. So far, the results are pretty good. Again, they are not as good as I would prefer.

Strange, but the MPEGs from my PVR "look" nicer. If I could edit MPEGs as nicely as DV encoded AVIs, I'd go that path. MSP6.5 is pretty good, but you still get that huge encoding time (blah) whenever fx and the like are added. I have still to find a good way to edit MPEGs.

What I really like about DV is that the files do not take a lot to capture, edit, or playback.

OK, enough ranting...