Video quality needs improving

[James-WBF]

Hey folks,

Been a while. I ran a thread a while back on building a new setup using SD video. The thread and solution was this:

churchmedia.net/forums/general-video-production/46108-software-hardware-encoding-solution.html (need to cut/paste - Links not allowed by me)

While it all worked fairly well, the quality of the video lacks when encoded. Actually, it seems to be good coming out of the video switcher, but when it hits the Grass Valley ADVC 110 and then via firewire into the workstation doing the encoding (Dell Precision 1500 w/Adobe CS5 On-location), the quality is lost..

Should I consider scrapping the ADVC 110 & firewire board, and get something direct into the PC (a Black Magic card or something)?

Thanks,

James

[sempei13]

How are you judging quality. On a computer monitor (with probably much greater resolution than 1024x768 ), 480i will look very small at native size and bad when blown up. If you output it to DVD and play on an SD tv, does it look alright?

Paul

[mjl5007]

To be clear, are you encoding in real-time, or just capturing in real-time and encoding in post? If the latter, I assume the captured video looks OK, and it's the final encoded video that lacks quality? What's your final output format?

Taking a shot in the dark here, but are you doing any deinterlacing as part of your encoding process? In my limited experience working with DV footage (which is interlaced) and exporting to a final progressive product (WMV, H.264 in MP4, etc), Adobe's deinterlacing routines are quite lacking... or at least I haven't been able to get them to give me any results that are remotely acceptable. I've been able to get much better results using AviSynth and the NNEDI2 and Yadifmod deinterlacing filters. Of course, the two problems with that are 1) it's much slower than real-time, even on modern hardware, and 2) it's certainly not a very user-friendly solution, and probably pretty difficult to integrate into a real-time live encoding workflow, even if they were fast enough.

So, if you're doing deinterlacing as part of your encoding process and you're using Adobe's deinterlacing filters, that might have something to do with it. Hard to tell without seeing a sample of the final product.

[James-WBF]

Sempei13,
Quality is judeged in two parts. 1), when I look at the video coming out-of the video mixer (DataVideo SE-500) which is showing up on the dual monitor output (7" screen), and that sits right next to the Encoding machine VGA monitor where it does not look as crisp (somewhat slightly blurred). 2) when viewed on the website, as told by the Church staff it does not look crisp (everyone's a critic .

mjl5007,
We are capturing real time, but not encoding at the same time. We start the encode process after the capture is complete. The captured video looks like the finished product, not very crisp. So, if the captured video is not good, then the final product cannot be improved upon.

The file size each week is about 500 Mb for a 43 minute video. To me personally, that also seems large, not to mention that's quite a download for folks wishing to watch it. Also, as can be seen, a secondary issue we are having is the video is not edge-to-edge, meaning that there is a left & right sides that have black space.

Here is a link to one such video.
waterbrook.org/sermons/Sermon_2011_10_30.wmv

As for the de-interlacing question, I would need to ask the video guy in the both that does this on a regular basis, but he may not be sure. It's been a big learning process to get this going. And of course, we're all volunteer's

Anyhow, thanks for the input.

James

[Gracetech]

500 MB isn't all that large but the good news is you can scale it back if need be. Flash is really the best encoder to maintain quality but loose alot of data.

I'm guessing you are loosing quality through your capture device or capture encoding.

You have a few issues going on.
1. Your footage is being squeezed horizontally from 16:9 to 4:3. You will need to configure either your capture our output settings to correct this.
2. You are capturing through alot of conversions. I'm gonna go out on a limb and say you probably have a digital processing switcher of which you are running into analog and out analog. Then to top it all off you go through another conversion when you hit your advc 110 and then go from analog back to digital again. These jumps through converters deteriorate your image on every conversion. If you are making some of these conversions on composite analog then it's going to be even worse.
3. You are over saturating the image. This is causing the colors to bloom which causes issues when going through conversions. This compounds the degradation of your image.

The best thing you can do at this point is to back up a bit and provide some details of your setup. You can keep it pretty simplistic.

Here is an example

Camera(UberCam 2000)
analog cable(composite signal)
Switcher(UberSwitcher 2000)
analog cable(Y/C...aka SVideo)
Capture Device(GV ADVC110)

Capture settings(UberCrazy codec)
Output encoding(UberSlimmer codec)

Screen captures and pictures are always welcomed since we may pick up on something you may not.

I'm sure the collective here can get your setup improved.

crt

[sien]

Hi James,
The Sony HDRFX1 has a component output, s-video output and composite output.
Which one are you using to send to the Datavideo SE500?

[James-WBF]

Gracetech,

I copied over the JPG showing the layout of how all the equipment is currently hooked up (It's also in my original post mentioned in my top posting) See attached.

I'll get the capture encoded specifics from the video guy this morning.

Sien,
We are using the Composite output if I remember correctly. I can call out there later this morning and have someone take a look as well.

Thanks All...

James

[mjl5007]

OK, forget the deinterlacing idea (for now).

To further add to Chad's point about the video being squeezed from 16:9 to 4:3, your final output video is at 720x480 which, without going into too many technical details, isn't a proper resolution for either 16:9 or 4:3 (barring anamorphic DVD video, which isn't necessary for web delivery). On top of that, those black pillar-boxes on either side of the video are being encoded into the video frame, which is nothing but wasted data; you're using up bits/bytes to encode empty black areas that you can get for free during playback. Once you get things reconfigured properly and get rid of those black pillar-boxes, there will be more bits available for encoding the actual video content at better quality (if you're keeping the same file size), or, you can get the same video quality with fewer bits (decreasing the file size). Either way it's an improvement.

On the topic of encoding settings and quality level vs. file size, here's a video of one of our sermons from several weeks ago for comparison (I also can't post links yet, so be sure to replace the [DOT] with a dot):

sermons[DOT]calvarysc.org/media/20111002video.wmv

Now, that video is only ~36 minutes long, and we're doing 4:3 output instead of 16:9, but the quality is acceptable (minus the improper deinterlacing resulting in combing when Pastor Dan moves around or waves his hands) and the file size is only ~139MB. So it's certainly possible to get your quality level up, but keep the file size more manageable.

[For the record, that video was recorded live-to-DVD via the composite output of a Sony DSP-PD170 into a Sony DVDirect, then ripped from the DVD and re-encoded to WMV using some one-click program that our IT/media staff person uses because it's simple. Quality could easily be improved if we used FireWire between the camera and DVDirect, and some better software for the transcode to WMV.]

[Gracetech]

Originally Posted by James-WBF

I copied over the JPG showing the layout of how all the equipment is currently hooked up (It's also in my original post mentioned in my top posting) See attached.

Sorry James, i was groggy when i posted and didn't read through everything.

720X480 is the resolution for DV. 640X480 is the resolution of SD...depending on who you talk to . Also for DV format it doesn't matter if it is 16:9 or 4:3 but that the clip is flagged to be in 16:9. This is the same for all of the SD world. Meaning that the video is the same resolution no matter how you stretch it. Your problem is that you have a 720X480 clip of which you are not taking proper usage of the resolution as MJL said. In essence your video resolution is around 480X480. Correcting that alone will greatly improve things. The next step is to upgrade your interconnection links as high as you can. This may mean using the humble S/V cable. If you can do component then that will really help.

Let us know about the encoding end of things and it's possible to gain some quality there as well.

Also while i'm thinking about it...i ususally will drop the frame rate to 24fps and ditch interlacing. This is not to get a cinematic effect but to improve the clarity of the video for the web. At 24fps less data will go towards extra frames and more to the quality of the frames. We are talking about micro adjustments here but every bit helps. Making the footage progressive also helps save on file size while keeping quality up. The reason is that keyframing works better on progressive footage than interlaced footage. Again micro adjustment.

crt

[James-WBF]

Thanks again for the input folks.

We seem to be having a bit of a challenge determining the "encoding method" and proper selection of the settings (maybe?).

Some things of note from the previous posts:

1) Oversaturating the image - How do I correct that?
2) Black pillar boxes - I need to reconfigure the encoding process to do 4:3 (or 480x480), and that should get rid of the pillar boxes?
3) Increase my video cable quality. So mabe go to an S-Video instead of composite.
4) Go to 24 fps. Software change I presume.
5) Make the footage progressive. This a software change?

What should I encode to? We are doing WMV & MOV (for those apple folks). Is this pretty much what everyone else is doing as well?

Thanks...

James

[mjl5007]

I'll respond to the items I can below...

Originally Posted by James-WBF

2) Black pillar boxes - I need to reconfigure the encoding process to do 4:3 (or 480x480), and that should get rid of the pillar boxes?

Well... yes and no. The issue here is that your camera records in 16:9, but you're outputting via composite and through the SE-500, which uses an NTSC signal at 4:3 only. So in order to get your widescreen image to the computer through a standard 4:3 system, the camera either has to squeeze the widescreen frame into a 4:3 frame, or letterbox it (black bars on top and bottom). The camera is probably configurable to do either, but don't quote me on that. The consequences of this are that you either have to re-stretch the squeezed widescreen video frame back out to 16:9 (853x480 if you're keeping the same frame height) once you capture it to make it look correct, or you crop off the black letterboxing after capture, but now you're working with a 640x360 widescreen frame, instead of 853x480; you've lost 43% of your resolution. Personally I'd go with the former (squeeze the widescreen image into a full 4:3 frame, then restretch it back out after capture). Of course, the gotcha there is that since you're cutting the slides in live, they also will take up the full 4:3 frame, but you *don't* want to stretch them back out to widescreen after capture because 4:3 is their correct ratio. So you'd have to cut your captured video at the points where a slide comes in or out, and only stretch the live video parts, but leave the slides in 4:3 pillarboxed.

At least, that's my take on the situation. Someone will correct me if I'm wrong.

Originally Posted by James-WBF

3) Increase my video cable quality. So mabe go to an S-Video instead of composite.

S-Video will help a bit, definitely, but don't expect miracles.

Originally Posted by James-WBF

4) Go to 24 fps. Software change I presume.
5) Make the footage progressive. This a software change?

Yes; since you're using an NTSC video signal which is 29.97 interlaced frames/sec, you'll have to change the framerate after capture. In addition, going to progressive footage means you *will* need to perform deinterlacing, which as I said before, I've never been impressed with Adobe's de-interlacing routines.

Originally Posted by James-WBF

What should I encode to? We are doing WMV & MOV (for those apple folks). Is this pretty much what everyone else is doing as well?

Depends on how you'd like to distribute. If you want folks to download the video files before watching, WMV and/or MOV works fine, although we just do WMV and let the Apple folks use Flip4Mac. If you want to embed them in players on your website and have them stream to the viewer in-browser, FLV is the way to go.

[mjl5007]

Thought about this a little more and realized I may have missed some important points...

I talked about setting the camera to output the full widescreen image but squeezed horizontally into a 4:3 frame, and then stretch it back out to 16:9 after capture. I checked the manual for your camera and that is indeed possible, but I neglected to consider that you're also displaying the program output of the SE-500 live on two widescreen TVs in the foyer and hallway, correct? If so, you're feeding a 4:3 signal to a widescreen display... are you having the TV pillarbox the incoming signal to keep it in the 'correct' aspect, or are you using the TV's 'WIDE' mode to stretch the incoming signal to fill the screen? I would guess that pillarboxing is what you want, since otherwise your 4:3 slides will look stretched out (although that may be acceptable to you).

If pillarboxing is what you're doing on the TVs, then having the camera feed a squeezed 16:9 image in a 4:3 frame is going to mean displaying that squeezed image at 4:3 on the TV's, which you don't want. What you probably need to do then (and maybe you're already doing this, although I didn't see it mentioned specifically anywhere) is to set the camera to record in DV mode (which is standard 4:3) instead of HDV (16:9). Then you'll be outputting an 4:3 image in a 4:3 frame, no squeezing or stretching required, all compatible with your 4:3 slides from the computer, the SD/4:3 switcher, cabling, etc. Then when you capture to the computer, you'll capture a 4:3 frame which won't require any kind of re-stretching, and you can export to a final video in 4:3 aspect.

On the other hand, if you're using WIDE mode on the TVs, and you're OK with your slides showing up stretched out (for as short a time as they're shown) as long as the video looks right, then my original suggestion in the previous post will work. Set the camera to record in widescreen HDV (IN/OUT REC Menu, REC FORMAT set to HDV1080i), and set the output TV type to 16:9 (IN/OUT REC menu, TV TYPE set to 16:9). The camera should then output the widescreen image squeezed horizontally into a 4:3 frame, so it'll look squeezed on your SE-500 monitors, but normal when you stretch it out using WIDE mode on the TVs. I'm not sure, but you may be able to get the capture to treat the incoming signal as widescreen off the bat, so you won't have any manual stretching to do in post (although your slides will be stretched out as they will be on the TVs), and you can just encode to a 16:9 output resolution (after setting things up for the conversion to 24fps and progressive).