Capturing in 352x240 or 640x480 for VCD format

[ELPOU]

HI, I have a AIW Radeon 7500 and I'm kind of happy of the results I get capturing in 352x240 Huffyuv and then converting the video to VCD with TMPGenc but I was wondering if capturing in 640x480 would give me more quality, any ideas !

ELPOU

[dcapp1]

I would like to know why your results capturing to 352x240 are acceptable because when I capture to 352x240, the result SUCKS! The only way I get good quality captures(same as source) is by capturing to 640x480 and then encoding to svcd with tmpgenc. Thats the only way.

[wulf109]

Do your 640x480 display correctly on your TV when played on standalone DVD player?? I thought 640x480 would not resize correctly on a TV.

[pyates]

For VCD format, capuring at the native res (352x240) is ideal. Capturing at ???x480 results in an interlaced resolution, which then has to be de-interlaced and re-sized by the mpeg encoder. All this would just take time, and it would not improve quality. In fact, it may reduce the quality. VCD's quality limitation has a lot more to do with the bitrate than it does with the resolution. For VCD, stick with your current process.

On the other hand, capturing at a higher res and encoding to SVCD is by far the better choice. When done correctly, the quality approaches that of DVD. Try this: Capture at 480x480 (not 640x480) with a good lossless codec like huffyuv or picvideo, then use VirtualDub to de-interlace and add the temporal smoother (4), then use DVD2SVCD (with CCE @ 4-pass) to encode to SVCD. The quality will be better than the original.

[pyates]

note: wulf109 is correct. 640x480 is not a standard res. As I mentioned, capture at 480x480 for the best SVCD results.

[pyates]

Sorry for being so chatty, but I am anticipating a flame about my claim that the SVCD will be better than the original. As an example; I use an ATI TV-Wonder VE to capture Smallville at 480x480 on regular analog cable TV. In the original, you can see typical analog artifacts like tiny flecks, etc. After my process, the SVCD looks like Digital Cable. It's just perfect, seriously.

[crahak]

My opinion: capture at 640x480, deinterlace and resize bicubic. Can't beat that quality-wise.

[pyates]

You really think a bicubic resize produces better quality than no resize at all?

Please explain.

[dcapp1]

My question again, How do you get good quality capturing at 352x240. I use the picvideo codec with a quality of 18, and the quality of the avi sucks. Im not saying that it cant be done, I'm just saying that whatever I'm doing isn't doing anything for my video.

[pyates]

Are you watching the avi at the original resolution or are you resizing it in your player, like to 200% or fullscreen or something. That would surely suck. But when played at 352x240 (100%), it should not suck at all. It is small, granted, but the quality should be fairly good. And remember, when it is displayed on your TV, it does not get "resized" like it does going fullscreen on your computer.

[dcapp1]

I used to test the video with Windows media player at 50%, and the video looked good. When I palyed it on my TV it sucked so I did some research and I found that zooming to 200% looks more like what it will on the TV. The capture zoomed to 200% sucks unless I capture at a higher resolution. I dont know, maybe I dont know what Im talking about.

[pyates]

The 200% thing is something I have never heard. I think you know what you are talking about, but my experience has not indicated what you have observed. If I capture at 352x240 and encode to VCD, the quality is typical VCD quality (which isn't great, admittedly, but what do you expect from 352x240 and 1150kbps). Maybe that's the whole issue, that VCD basically sucks compared to SVCD?

I really don't think capturing to 640x480 and then encoding it down to 352x240 @ 1150kbps will give any better result.

[pyates]

dcap1, are you using TV-out from your video card or a standalone deck to play the VCDs?

[dcapp1]

I am using a standalone deck to play the vcds

[pyates]

hmmm, and they suck _that_ bad? I think it must be just the typical VCD suckiness.

Try out the temporal smoother and de-interlacer on your SVCDs, though. It's pretty sweet...

[dcapp1]

I'm sorry you are misunderstanding me, After I capture to 640x480 I encode to svcd. I used to capture at 352x240 and encode to vcd, and the quality was bad. VCD quality isn't that good to begin with, but my final product was worse than vcd quality.

[dcapp1]

Thanks, I will try your method, but when I add both of those filters and save as new avi Vdub will tell me I am saving uncompressed avi and the file will be very large. What shall I do about this because I want to keep the file at its original size after capture.

[hwoodwar]

dcapp1:

You can turn back on compression before you save the filtered file by going to Video>>Compression and selecting the compression codec (probably HuffyUV) and then going to save the AVI.

[jeremypps]

looking in the topics of this forum, i've read several times it's good choice, when trying to make a vcd, capturing at 352x576 (i work with pal res, so 320x480 for ntsc) in virtualdub using vertical reduction, in this way you capture both fields and no need to deinterlace. the problem is i still don't know if it's better bilinear or bicubic.
the other problem is the recompression, you capture a compressed avi and then recompress it again and then convert (and recompress) to mpeg. isn't it better do the filtering with tmpgenc?
finally my two cents: svcd is theorically better than vcd, BUT if you have not a good source AND if you want good compromise between size and quality AND thinking at the fact that mpeg2 gives his best at bitrates higher than the ones used for svcd, well, vcd is not always a bad choice.

[86peaches]

why would you want to de-interlace if you're making (s)vcd's to playback on a stand alone player?

[PlaiBoi]

Or, why dont you just use PowerVCR and capture directly to (S)VCD mpeg-1/2? I've heard that PowerVCR gives promising results. Any ideas?

[Nelson37]

Well, here's my two cents.

If you cap at half-res, you are throwing away one full field of video data, which is lost. Cap at 480 (576) and resize to 240 arrives at VCD res using all available video to get there - but takes time, usually looks better.

Vertical res is where data may be LOST, horizontal res it may be underrepresented (?), but it's all there.

The pixel variances and different resolution and sharpness make comparing PC and TV only good for rough comparison. 640 x 480 is decent approximation but try cropping to 672x448 and adding borders later. Rather than capping less of the whole picture I cut off unseen parts of the whole first, if that makes sense.

Full-screen display, on 15" anything looks good, 17" with no or very few blocks looks good on 27" TV, if it looks sharp and crisp with no blocks on my 19" monitor it looks DVD-like on the TV.

Did a LOT of comparisons between 352x480, 480x480, and 720x480, with resize to 480x480 and 352x480. Slight difference, took multiple tests to confirm, but larger cap resized looks better to me. Takes longer, though.

I don't use filters unless cap is broadcast rather than digital, difference is so major I rarely cap from broadcast. Once smoothers come into play, VCD also becomes a choice as that's the only way I've ever made a decent VCD. CVD and SVCD with no filters, Digital source, looks as good as original broadcast.

[vitualis]

As what Nelson37 said eloquently.

@ pyates: capturing at full vertical resolution makes a difference. IMHO, it can make a BIG difference. It is the difference between capturing only one field per frame and both fields. With capturing one field, you are essentially throwing away up to half the video data.

Yes, you can get interlacing (which is why de-interlace may be required), but the interlacing tells you that you are getting video data you will otherwise have just thrown out. Generally, by capturing the full vertical resolution, the picture should be noticeably sharper and there should be noticeably less video noise.

Regards.

[pyates]

Or, why dont you just use PowerVCR and capture directly to (S)VCD mpeg-1/2? I've heard that PowerVCR gives promising results. Any ideas?

I dump my Smallville SVCDs onto a server at work so the night shift can watch them on their workstations (as well, I play them at home on my Apex). I have heard that superior standalone quality would result from not de-interlacing. Is this true?

[pyates]

@ Nelson37 and virtualis: I stand corrected. You make very valid points. In theory, anyway. I have done some testing of my own and I find it very difficult to tell the difference between a VCD that was converted from a 480 vertical res and one that was converted from a 240 vertical res. Once they are played on my TV through my standalone player, they are very, very similar. Not different enough to justify the extra encoding time.

In the end, if my source is 480 vertical, then I just go ahead and make an SVCD, since I find the quality to be consistently, significantly better.

As far as the question about re-saving the file in VirtualDub after applying the filters; don't re-save it at all, frameserve to your mpeg encoder. It saves HD space and saves a re-encode, which preserves quality. I just learned how to do this via the howto here at VCDHelp. It took me a while because I was missing MSVCR70.dll but I am glad I made the effort to finally figure it all out. It is really sweet. I also found some really nice add-on filters for VirtualDub, like the Subtitler and some nice Athlon optimized smoothers, etc.

Check it all out. I'm glad I did.

Regards to all,

P