Resampling quality setting in Ulead VideoStudio

[eclipse95]

In the "edit" tab of the preferences settings for Ulead VideoStudio is a setting called "resampling quality." The selections are good, better and best. The manual for the program does not specify what this setting does; it only says that it affects the quality of the video. Not very helpful. The setting does NOT relate to the MPEG encoding process. I have been unable to identify (including on the Corel fourms) what this setting actually affects/does. Does anyone know?

[jagabo]

If it involves the video it probably refers to the resizing filter used to shrink or enlarge frames. Take a small video, say 480x360, and enlarge it to 720x480. Look carefully at the resulting frame. Different resizing filters result in different sharpness and different artifacts.

480x360 test pattern:


720x480 Point (Nearest Neighbor) enlargement, ugly:


720x480 Precise Bilinear enlargement, few artifacts but not very sharp:


720x480 Precise Bicubic enlargement, few artifacts, sharper, but some oversharpening halos:

[zoobie]

I set all settings to best
strange the VS guys don't know

[jagabo]

I set all settings to best
strange the VS guys don't know

Simply selecting "best" isn't always going to be best. For example, using a sharpening resizing filter can cause over sharpening halos with material that's already sharp. If your source video has macroblock artifacts it will make them worse. And it will increase the bitrate requirement because it accentuates noise. I would do some experiments to see exactly what each of the settings does then use whichever one is most appropriate for each particular video.

[eclipse95]

Thanks. I had suspected that it might have something to do with if the resolution was being changed when saving the file, but I'm not so sure. The reason is that I took a short (1 min 45 sec.) DV clip, opened it in VideoStudio, and re-saved it as a new video file, making no changes to resolution, file type, and with no edits, filters or transitions. One would expect that under those conditions, the file would be created almost instantly. Nonetheless, it took VideoStudio between 20 and 45 seconds to create the file, with the resampling set to "good" taking the least amount of time, and set to "best" taking the longest amount. Obviously, the setting is affecting something. I'm just trying to figure out what.

[jagabo]

I took a short (1 min 45 sec.) DV clip, opened it in VideoStudio, and re-saved it as a new video file, making no changes to resolution, file type, and with no edits, filters or transitions. One would expect that under those conditions, the file would be created almost instantly. Nonetheless, it took VideoStudio between 20 and 45 seconds to create the file, with the resampling set to "good" taking the least amount of time, and set to "best" taking the longest amount

So your output was DV AVI too? Since you weren't resizing (unless you converted 16:9 to 4:3 or vice versa) that would imply the resampling option isn't realted to the resizing filter. Given the conversion time, I suspect it controls some setting in the DV encoder.

[eclipse95]

Thanks. Yes, it was output as DV AVI, same as the original, with no change in aspect ratio or anything else. Amazing that there appears to be no info as to the actual function of this setting. And in my example, one would think there is nothing to resample. Strange. The reason I raised this issue in the first instance is because I took my analog home videos (over 20 hours of Hi8), converted them to DV AVI, and, in VideoStudio, cut out some garbage and added transitions between scenes (with no other editing), then saved them as DV-AVI for archival purposes. But in doing so, I never noticed that the resampling quality was set to "better" rather than "best." And it's paramount that these videos be archived in as high a quality as possible. I sure don't want to have to reconvert all the original tapes again if it isn't going to make a difference. I've done test clips in both better and best settings, and my untrained eye doesn't see a difference on the monitor, but that doesn't mean there isn't some subtle yet perceptible difference between the two. That's why I'm trying to determine just exactly what this setting does.

[lordsmurf]

Fractal method:
Not as good on the straight lines, but superior on non-linears.

[eclipse95]

Not to be ignorant, but will that pattern help resolve the issue? I'm not changing resolution in the videos. Both original source and output are 720 x 480.

[jagabo]

No, smurf's sample won't help you since your issue doesn't appear to involve resizing. He's just on a fractal kick.

Post short samples of each and I'll take a look. Look closely at 4x enlargements. Look at smooth gradients, areas where there are lots of small details and/or noise, sharp edges, places like that.

[lordsmurf]

He's just on a fractal kick..

mmmmmmmmmmmmmmmmaybe. :P :P :P

[eclipse95]

I would post short clips (REALLY short since we're talking about DV AVI), but I don't see a way to post anything other than images.

[creamyhorror]

I would post short clips (REALLY short since we're talking about DV AVI), but I don't see a way to post anything other than images.

Upload them to mediafire.com and post the URLs here.

[jagabo]

Just use the upload image/file option below the edit box. You don't specifically have to mark it in the text.

[eclipse95]

Here's the URL to two samples, one in "better" and one in "best": http://www.mediafire.com/?sharekey=82fd1c9881b787f54c17ca8801618ef7e04e75f6e8ebb871

[poisondeathray]

Here's the URL to two samples, one in "better" and one in "best": http://www.mediafire.com/?sharekey=82fd1c9881b787f54c17ca8801618ef7e04e75f6e8ebb871

A and B are identical , you can verify this by using difference masks

Perhaps the setting has to do with rendered sections like transitions, titles, overlays etc..(of which there were none of in the samples) , not sections that were untouched ...

OT: that fractal uprez isn't very good IMO

[jagabo]

Those video frames in those two clips were identical. I used an AviSynth script that subtracts one image from the other then amplifies the differences. But there were none. Here's what it looks like when there are differences (two different frames are being compared here, frame 39 in one video, frame 40 in the other, nearly no motion between the two frames, images half size for posting):



And here's what comparing the two videos gave:



The big expanses of gray at the bottom mean the frames are identical.

[eclipse95]

Thanks for taking the time to review and analyze these. And thanks for providing the answer that I wanted to hear.

[eclipse95]

One final question -- would the above analysis have reflected differences in all parameters, including brightness? I could swear that the one on the right looks a little brighter (e.g., the boy's hair color, the dark portions of the water, the grass on the bank behind the water). But one would think that "resampling" wouldn't affect that. Do you recall whether the A and B listed in the corners of your posted frames of the two different videos correspond to the "A" and "B" of the samples I posted?

[jagabo]

If every pixel had been N units lighter or darker you would be able to tell by the shade of gray in the differences images. The 128,128,128 (R,G,B) in the comparison indicates there was zero difference.

Yes, A corresponded to Sample A, and B corresponded to Sample B. I just made the names shorter so it would be easier to type them into the Avisynth script.

Note that the term resampling often refers to audio. For example when it is resampled from 44,100 Hz to 48,000 Hz. And there are often different options available, some better (and slower) than others.

[eclipse95]

Okay. Thanks again.