Down converting HD from the editor at DVinfo.net

Pete Cofrancesco · May 26th, 2009, 02:43 PM

I shot a talent show with a Sony Z1u in HD, imported it into Final Cut Pro as HD and used Compressor to down convert it to Anamorphic SD. Looks good in the editor but when I watched the DVD on my wide screen tv, the detail on the faces and anything moving looks pixelated. These areas are where the interlacing is most visible. When I played the original tape through my camera out to a s-video cable to my wide screen tv it looks much better.

I've read other threads where people have had good results down converting after editing but for some reason I'm losing quality. I'm currently applying a deinterlace filter to the entire movie but it takes a very long time to render then down convert (like 9 hrs), doesn't seem worth all the extra time. I get the feeling that flat screen tvs might just like interpreting the analog signal directly from the camera rather than deinterlacing footage from a digital signal of the dvd. I wonder if down converting while importing from the camera would yield better results. I also wonder if it would look better if I filmed with a progressive camera. I find myself spending a lot of time deinterlacing because most of my movies are viewed on non interlaced viewers such as flat screen tvs or computer screens for web distribution.

Robert Lane · May 26th, 2009, 08:20 PM

Just to be clear, you always lose quality when down-converting from HD to SD because you're literally throwing away information in order to squeeze the data rate down to what a DVD player can handle.

Making a good looking MPEG2 encode for DVD is easily done but also a time consuming process; without knowing how long your project is it's impossible to say if 9 hours is "too long" or in fact on-target. That also depends highly on the settings you're using in Compressor.

I'd suggest two things: One, take a look at this thread where I've shared optimal settings in Compressor for good-looking DVD encodes:

http://www.dvinfo.net/conf/non-linea...artifacts.html

Two: As stated in that thread pick up the book, "Compressor 3 Quick Reference Guide, Brian Gary", and get a full understanding of the inside tricks into how Compressor and MPEG2 encoding really work.

Pete Cofrancesco · May 26th, 2009, 11:33 PM

I figured it out, the problem resides in how my flat screen TV up scales the video. I took some still photos and realized that for the dvd it stretches the the image to fill the screen while from the camcorder it letter boxes it and results in a nicer up scaled image. I assumed that wide screen tvs would be 16x9 ratio well I guess not. My tv is wider than that so even though the video was shot 16x9 it still either has to stretch or letter box it and for whatever reason straight from the camera the image looks cleaner with less artifacts. I'm pretty sure it doesn't have anything to do with how the video was saved as a mpg since I'm using the highest quality bitrate.

If you look at the girl on the podium you can see the difference. In person its easier to see the quality difference.

http://i14.photobucket.com/albums/a3...est/camera.jpg
http://i14.photobucket.com/albums/a3...n/test/dvd.jpg

The 9 hrs is the time it takes to render a interlace filter in fcp on a 45 minute movie. On a dual G5 renders take forever because its in HD.

Robert Lane · May 27th, 2009, 08:47 AM

If you down-scaled your footage to SD anamorphic as you stated may be part of the problem. "Anamorphic" was used primarily for SD-video or 16mm cameras using anamorphic lenses which would strectch the image vertically for later streching to fill a 16x9 or academy-frame size. Then either in editing or during MPEG encoding there would be an anamorphic "filter" applied to the footage to widen the image and correct for the vertical distortion.

DVD players look for flags built into the MPEG2 layer that tell it whether or not it's "full screen" or if it needs to stretch the image to fill. Those flags are created in the MPEG2 encode and are transferred to the authoring application.

Make sure your DVD player isn't set to force an aspect ratio change. In the future your HD projects should be encoded for SD-widescreen, not anamorphic. Then you'll get the letterbox bars on top and bottom and the DVD player won't try to stretch the image.

Liam Hall · May 27th, 2009, 10:06 AM

Quote:

Originally Posted by Pete Cofrancesco

I assumed that wide screen tvs would be 16x9 ratio well I guess not.

Except for the new 2:1 Phillips TV, they are 16:9. Your image looks pixelated because you've thrown away so much data in creating a 4:3 letterbox version which is then upscaled. You need to create an SD FHA (Full Height Anamorphic) master.

Quote:

Originally Posted by Robert Lane

"Anamorphic" was used primarily for SD-video or 16mm cameras using anamorphic lenses which would strectch the image vertically for later streching to fill a 16x9 or academy-frame size. Then either in editing or during MPEG encoding there would be an anamorphic "filter" applied to the footage to widen the image and correct for the vertical distortion.

DVD players look for flags built into the MPEG2 layer that tell it whether or not it's "full screen" or if it needs to stretch the image to fill. Those flags are created in the MPEG2 encode and are transferred to the authoring application.

Anamorphic lenses SQUEEZE the image horizontally, they don't STRETCH it.

Widescreen TVs work along the same principle to horizontally decompress a 4:3 anamorphic video image to create an image with a 16:9 aspect ratio.

Quote:

Originally Posted by Robert Lane

In the future your HD projects should be encoded for SD-widescreen, not anamorphic. Then you'll get the letterbox bars on top and bottom and the DVD player won't try to stretch the image.

No, you are simply throwing away scan lines this way. Much better to create an anamorphic master with the correct 'flags' for the DVD/TV set-up.

Pete Cofrancesco · May 28th, 2009, 12:02 AM

Quote:

Originally Posted by Liam Hall

Except for the new 2:1 Phillips TV, they are 16:9. Your image looks pixelated because you've thrown away so much data in creating a 4:3 letterbox version which is then upscaled. You need to create an SD FHA (Full Height Anamorphic) master.

Its rather a confusing issue and there are a number of independent variables. Whether its the TV, the DVD player, or the encode/down convert at the computer that is the culprit. Needless to say that when SD is upscaled on a HDTVs it can look bad. In my case not the entire image but edges and small details in certain areas are blocky or jagged.
I found a thread on another board about a similar issue so I know I'm not crazy.
Upscaling SD to HD on DVD for HDTV ?

I understand the point if artifacts are introduced during the encoding then they will be magnified when upscaled on a large screen TV. But in my case I don't thinks so because I'm using the highest bit rate. Although I'm running a test to deinterlacing it in the editor then encoding with two pass. My reasoning it will soften the problem areas and upscale smother.

To correct what I early said about my TV being wider than 16x9, I actually think it is 16x9 just that how these TVs upscale isn't as intuitive as one would think. And actually when viewed closely at the two screen shots I posted it's hard to say which if either kept the correct proportions. But I'm now sure when viewed directly from the camera it's letterboxed, thus maintaining its true proportions and upscaling with higher image quality, but at the cost of not filling the screen.

Here's an interesting article on anamorphic. Anamorphic DVD's Explained

One thing I wrongly assumed that when I shot video in widescreen mode it was using a different pixel ratio (more width wise and less height wise, but come to find out its the same amount of pixels just squished horizontally then expanded by the dvd player.

Pete Cofrancesco · May 28th, 2009, 08:14 PM

To wrap things up. Applying a deinterlace filter and using 2-pass encode yielded better results, but it took 24 hours of encoding and rendering! Not a process I could live with on regular basis.

I also noticed that when Anamorphic is used, the dvd player upscales the video larger than the TV screen and then crops about the same amount it would do for a CRT TV. I'd imagine that setting it to letterbox in DVD Studio Pro would yield better image although it wouldn't fill the screen.

Robert Lane · May 29th, 2009, 07:27 AM

Pete,

Pick up the book I recommended; it will very quickly and distinctly clarify all the issues you're scratching your head over.

Pete Cofrancesco · May 29th, 2009, 10:14 AM

Quote:

Originally Posted by Robert Lane

Pete,

Pick up the book I recommended; it will very quickly and distinctly clarify all the issues you're scratching your head over.

thx when I get some time I'll read the info detailed in the thread and check out the book. I've read articles on maximizing compressor encode, like using best or better motion estimator. That setting really increased the encode time so I usually leave it at the default setting. One of the things that Compressor can't do is control the amount of deinterlacing. For this you really have to use a deinterlaced filter in fcp. I've also read about Magic Bullet makes something that softens the edges of interlaced motion areas.

Pete Cofrancesco · June 6th, 2009, 12:14 PM

I finally figured out the problem. In Compressor I was using the Deinterlace filter. When it is set to Even or Odd, it creates the blocky noisy grainy look I was complaining about, when it is set to Blur it makes everything look out of focus.

So never use the deinterlace filter in Compressor, its really best to leave it interlacing even if your viewing in on a progressive device such as a flat panel tv.

May 26th, 2009, 02:43 PM	#1
Pete Cofrancesco Inner Circle Join Date: May 2004 Location: USA Posts: 3,005	Down converting HD from the editor I shot a talent show with a Sony Z1u in HD, imported it into Final Cut Pro as HD and used Compressor to down convert it to Anamorphic SD. Looks good in the editor but when I watched the DVD on my wide screen tv, the detail on the faces and anything moving looks pixelated. These areas are where the interlacing is most visible. When I played the original tape through my camera out to a s-video cable to my wide screen tv it looks much better. I've read other threads where people have had good results down converting after editing but for some reason I'm losing quality. I'm currently applying a deinterlace filter to the entire movie but it takes a very long time to render then down convert (like 9 hrs), doesn't seem worth all the extra time. I get the feeling that flat screen tvs might just like interpreting the analog signal directly from the camera rather than deinterlacing footage from a digital signal of the dvd. I wonder if down converting while importing from the camera would yield better results. I also wonder if it would look better if I filmed with a progressive camera. I find myself spending a lot of time deinterlacing because most of my movies are viewed on non interlaced viewers such as flat screen tvs or computer screens for web distribution.

May 26th, 2009, 08:20 PM	#2
Robert Lane Go Go Godzilla Join Date: Nov 2004 Location: Scottsdale, AZ USA Posts: 2,823 Images: 15	Just to be clear, you always lose quality when down-converting from HD to SD because you're literally throwing away information in order to squeeze the data rate down to what a DVD player can handle. Making a good looking MPEG2 encode for DVD is easily done but also a time consuming process; without knowing how long your project is it's impossible to say if 9 hours is "too long" or in fact on-target. That also depends highly on the settings you're using in Compressor. I'd suggest two things: One, take a look at this thread where I've shared optimal settings in Compressor for good-looking DVD encodes: http://www.dvinfo.net/conf/non-linea...artifacts.html Two: As stated in that thread pick up the book, "Compressor 3 Quick Reference Guide, Brian Gary", and get a full understanding of the inside tricks into how Compressor and MPEG2 encoding really work.

May 26th, 2009, 11:33 PM	#3
Pete Cofrancesco Inner Circle Join Date: May 2004 Location: USA Posts: 3,005	I figured it out, the problem resides in how my flat screen TV up scales the video. I took some still photos and realized that for the dvd it stretches the the image to fill the screen while from the camcorder it letter boxes it and results in a nicer up scaled image. I assumed that wide screen tvs would be 16x9 ratio well I guess not. My tv is wider than that so even though the video was shot 16x9 it still either has to stretch or letter box it and for whatever reason straight from the camera the image looks cleaner with less artifacts. I'm pretty sure it doesn't have anything to do with how the video was saved as a mpg since I'm using the highest quality bitrate. If you look at the girl on the podium you can see the difference. In person its easier to see the quality difference. http://i14.photobucket.com/albums/a3...est/camera.jpg http://i14.photobucket.com/albums/a3...n/test/dvd.jpg The 9 hrs is the time it takes to render a interlace filter in fcp on a 45 minute movie. On a dual G5 renders take forever because its in HD.

May 27th, 2009, 08:47 AM	#4
Robert Lane Go Go Godzilla Join Date: Nov 2004 Location: Scottsdale, AZ USA Posts: 2,823 Images: 15	If you down-scaled your footage to SD anamorphic as you stated may be part of the problem. "Anamorphic" was used primarily for SD-video or 16mm cameras using anamorphic lenses which would strectch the image vertically for later streching to fill a 16x9 or academy-frame size. Then either in editing or during MPEG encoding there would be an anamorphic "filter" applied to the footage to widen the image and correct for the vertical distortion. DVD players look for flags built into the MPEG2 layer that tell it whether or not it's "full screen" or if it needs to stretch the image to fill. Those flags are created in the MPEG2 encode and are transferred to the authoring application. Make sure your DVD player isn't set to force an aspect ratio change. In the future your HD projects should be encoded for SD-widescreen, not anamorphic. Then you'll get the letterbox bars on top and bottom and the DVD player won't try to stretch the image.

May 28th, 2009, 08:14 PM	#7
Pete Cofrancesco Inner Circle Join Date: May 2004 Location: USA Posts: 3,005	To wrap things up. Applying a deinterlace filter and using 2-pass encode yielded better results, but it took 24 hours of encoding and rendering! Not a process I could live with on regular basis. I also noticed that when Anamorphic is used, the dvd player upscales the video larger than the TV screen and then crops about the same amount it would do for a CRT TV. I'd imagine that setting it to letterbox in DVD Studio Pro would yield better image although it wouldn't fill the screen.

May 29th, 2009, 07:27 AM	#8
Robert Lane Go Go Godzilla Join Date: Nov 2004 Location: Scottsdale, AZ USA Posts: 2,823 Images: 15	Pete, Pick up the book I recommended; it will very quickly and distinctly clarify all the issues you're scratching your head over.

June 6th, 2009, 12:14 PM	#10
Pete Cofrancesco Inner Circle Join Date: May 2004 Location: USA Posts: 3,005	I finally figured out the problem. In Compressor I was using the Deinterlace filter. When it is set to Even or Odd, it creates the blocky noisy grainy look I was complaining about, when it is set to Blur it makes everything look out of focus. So never use the deinterlace filter in Compressor, its really best to leave it interlacing even if your viewing in on a progressive device such as a flat panel tv.