Whats the benefits of half D1 ?

[jesse40902]

I heard half D1 will give u a better picture quality if used right? Also if using half D1, is it necessay to do a 2 line pass to encode as well?

[mats.hogberg]

Not necessarily, but at lower bitrates (let's say <3000 kbps), it will be better than full D1. 2 (or more) pass VBR usually provides better quality at lower bitrates too, but necessary, not.

/Mats

[edDV]

D1 resolution is 704 to 720x480. Half D1 resolution is 352x480. In theory half D1 requires half the datarate to yield a similar ecoding quality.

Source quality for VHS and OTA NTSC fall at or below 352x480. In those cases you should get similar image quality if you encode to half D1*. DTS and cable channels have a source resolution in the 480-540x480 range. In those cases half D1 will yield a slight reduction in horizontal resolution.


* The actual quality at half D1 depends on the technology of the capture card.

[greymalkin]

* The actual quality at half D1 depends on the technology of the capture card.

that's a good point to make. many times people differ on whether or not 352 x 480 looks much different than the vhs/cable source.. it usually sounds like this

greymalkin: 352 x 480 looks just fine with vhs/analog cable

someone else: you need your eyes checked man. that res looks like total garbage and even my dead grandmother could see the difference. If your grandmother is dead I should go spit on her grave for having such a stupid grandchild.

greymalkin: hmm..i never could tell a difference before..i guess i'm just blind and should go shoot myself immediately.

[edDV]

General rule for maximizing quality at half D1 is capture at the native resolution of the capture card (high bitrate) and then MPeg2 encode to 352x480/576 (VBR to fit).

Some older Connexant cards are 640x480 native. Many of the newer ones are 704 or 720x480/576 native.

I've never heard of a card that captures natively to 352x480/576 without scaling. Realtime scaling is usually very crude.

Hauppage PVR cards have a built in hardware MPeg2 encoder tthat can be set to either 720x480/576 or 352x480/576 output.

The ATI 550 tuners capture and hardware MPeg2 encode to 720x480/576 only.

[sacajaweeda]

To the OP:

Here's a bit of an analogy for you that I like to use when trying to get somebody to grasp the whole frame size & bitrate concept.

Think of it like this. You are a painter. The section of a wall you are painting is your frame. You have just one can of paint that you can use to paint it. The paint is your bitrate. You can stretch it out with a thinner coat and paint the entire section of wall or you can paint an area half the size with the same amount of paint. One will look pretty good but perhaps a little spotty here and there, and the other will have a much more consistent look because you weren't stretching it so thin to cover twice as much space.

HTH

[TheFamilyMan]

For all the VHS movies I've backed up to DVD, I captured at 640x480 and then encoded to 352x480 using a variable bit rate maxed at 6000. I can fit 2-3 hours on a single DVD this way. The results are as good as the original, (possible better due to noise reduction?) though I don't play them back on a huge projection system or a HD ready TV. This is where 1/2 D1 shines IMHO.

Edit: For good quality VHS to DVD conversions, video noise reduction is absolutely necessary!!!!

[shardison]

I've tried half D1 in many situations that seemed to call for it, but I don't think it ever looked better than plain old full D1, even at lower bit rates. I don't like the loss of detail, even at the expense of a few extra blocks showing up. Probably my software, but maybe not. Below 3000kbs, full D1 starts to really suffer, and switching to half D1 doesn't seem to help. Below 3000kbs you might as well go 352x240 and get a nice, clean compression ( a bit blurred, but easy on the eyes). At least that's what I've decided after years of futzing around.

[GMaq]

Hi,
Something else to consider is if you are going to do anything with this video besides put it on a DVD, I've been doing a lot of conversions to MPEG-4 lately for iPod use and all the stuff I did in half D1 ain't looking so pretty once it's been encoded @ 512x384 whereas full D1 from the same source looks better because I'm downsizing the resolution instead of vice versa, Six months ago I wouldn't have thought it'd make any difference. Anyway no format is gonna last forever so if you think you may ever be converting to a different format (Whatever it may be ) it may be worth using full D1 with an appropriate bitrate even if it exceeds your requirements for the source. Just my $.02!

[edDV]

Hi,
Something else to consider is if you are going to do anything with this video besides put it on a DVD, I've been doing a lot of conversions to MPEG-4 lately for iPod use and all the stuff I did in half D1 ain't looking so pretty once it's been encoded @ 512x384 whereas full D1 from the same source looks better because I'm downsizing the resolution instead of vice versa, Six months ago I wouldn't have thought it'd make any difference. Anyway no format is gonna last forever so if you think you may ever be converting to a different format (Whatever it may be ) it may be worth using full D1 with an appropriate bitrate even if it exceeds your requirements for the source. Just my $.02!

Very true. Compression quality depends first on the quality of the source. Just look at the quality of a VCD or SVCD encode from a HDTV MPeg2_TS master.

[davideck]

I've tried half D1 in many situations that seemed to call for it, but I don't think it ever looked better than plain old full D1, even at lower bit rates. I don't like the loss of detail, even at the expense of a few extra blocks showing up.

This has also been my experience.
With the Hauppauge 250, I always capture at 720.

In theory half D1 requires half the datarate to yield a similar ecoding quality.

Correct me if I'm wrong, but I would argue the following;

I think it is necessary to distinguish between the source image resolution (the amount of detail in the source image), the sampling resolution (capturing at 352 or 720 pixels per line), and the captured image resolution (the amount of detail preserved in the captured image).

I think that the primary factors which determine the bitrate requirements are;

(1) The captured image resolution
(2) The amount of motion in the captured image
(3) The amount of noise in the captured image, because it disguises as (1) and/or (2)
(4) The amount of timebase error in the captured image, because it also disguises as (1) and/or (2)

I think that the sampling resolution is only a secondary factor whose affects are determined by each particular capture/encoder implementation.

Since all of these factors must share the allotted bitrate, it can be quite an interactive mess.

In addition, there are multiple compression techniques incorporated into MPEG_2 encoding, each exploiting some type of image and/or data redundancy.

Consider a flat field; one uniform color that fills the entire raster.
An ideal encoder would compress a flat field captured at 720 almost as well one captured at 352. The bitrate requirement for both captures would be about the same. Even with low resolution sources, the bitrate requirement for both captures would be about the same.

But if the source resolution exceeds that which can be preserved with a 352 sampling resolution, then the bitrate requirement for the 720 capture will exceed that of the 352 capture. This is because the 352 sampling resolution reduces the captured image resolution, thereby reducing the bitrate requirement.

Of course, noise plays a big part in this. Capturing a noisy source at 720 might only preserve more noise at the expense of bitrate; basically a waste. The same is true for timebase errors.

But for clean and timebase stable source images, a 720 capture should only require a significantly higher bitrate than a 352 capture if it is actually preserving more detail.

Whether or not a particular capture/encoding implementation exhibits this behavior is up for grabs. YMMV.

[edDV]

I would agree with most of that. I was thinking of a typical off cable somewhat noisy analog capture at 720x480 and then encoding that to either 720x480 or 352x480. The source horizontal resolution would range from ~352 (analog NTSC) to ~704 (for downscaled HD or SD DTV). Such sources at least have a stable timebase. Timebase jitter from a vcr often gets detected as motion by MPeg encoders.

I would think a 3-3.5MHz low pass luminance filter applied before encoding would remove higher frequency detail + noise making for more efficient encoding to 352x480.