Bilinear MMX vs. Bicubic

[Eric#]

Hi there,

i have a question regarding the RESIZE-options of DVDx 2.0.
What's the difference between BILINEAR MMX and BICUBIC?
The only difference *I* did figure out was that bilinear is a much faster (12-14 fps encoding) than bicubic (4-6 fps).
I couldn't see any difference in the quality of the pictures.

So what's the deal about BICUBIC resizing?

But i have to say that i was ripping a DVD from 720x576 to 720x384, so i just cut away the black bars, but didn't really RESIZE the picture. Maybe therefore i saw no difference. But still - the encoding time drastically decreased while using bicubic.

Can someone explain me why - and what's the advantage of bicubic resizing (i guess when it's more than 2x slower there should be an advantage )

regards

[snowmoon]

Resizing images is a comples science.

If you are only resizing on one axis then bi-linear should be identical to bi-cubic. Most algorithms are more importand for scaling images UP not scaling them down.

[sterno]

I seem to remember that the avisynth documentation said that bicubic usually gives better results for enlarging the image, but for reduction bilinear is just as good.

[PigOnWing]

A while ago I stumbled upon a comparison test which was invaluable when making my final decisions.

As it says, the general notions is 'Use Bicubic for upscaling, bilinear for downscaling' - and Avery Lee's VDub's documentation states that when using Bilinear shrinking, it can effectively 'squash noise'. I read somewhere a quote from Avery Lee, that on a general notion, he uses precise bicubic for upsizing, precise bilinear of downsizing a pretty good source, and precize bicubic downsizing if the source is more problematic.


-- Piggie

[Eric#]

Thanks guys for explanation and the helpful links

[bbb]

Always use bicubic for upsizing and downsizing.

The purpose of using bilinear to washout noise is unsatisfactory. What bilinear downsizing does is take out detail, sharpness, and accurate color definition. Just look at the examples. There is no reason to use bilinear resizing. The captions tell you it's better for downsizing but does it really look better? No.

Bicubic preserves detail and sharpness. Currently, bicubic is the best for any type of resizing. In the future, better resizing algorhythms may be adapted to VirtualDub filters such as B-Spline and S-Spline. For now, just stick with bicubic.

[PigOnWing]

The purpose of using bilinear to washout noise is unsatisfactory. What bilinear downsizing does is take out detail, sharpness, and accurate color definition. Just look at the examples. There is no reason to use bilinear resizing. The captions tell you it's better for downsizing but does it really look better? No.

I guess it's a matter of taste. Personally, I like sharpening images - many people like softening them.
Bilinear downsizing is a bit 'softer'. Bicubic one, though indeed preserving more detail, also preserves and actually increases pixelation, in some places (and personally, I think the examples indeed do show this clearly).

'Bicubic for Upsizing, Bilinear for downsizing' is a General Rule-of-thumb. It's a rule-of-thumb which, eventaulyl, both deals and has its origins with matters concerning to taste.

You don't like that rule-of-thumb and prefer another one? Don't use it and use the other one. Nobody will get angry.

-- Piggie

[bbb]

Yes, it's a matter of taste. People with higher taste for better quality video and patience use bicubic, deinterlace their video, and use CVDs & SVCDs over VCDs.

People with poor taste will stick with bilinear, don't deinterlace, and stick with VCDs for the sake of speed and don't really care about quality.

Such is the nature of most people in this world who just don't give a damn about the quality of their work. This lazy attitude, however, is nothing to aspire towards.

[snowmoon]

Yes, it's a matter of taste. People with higher taste for better quality video and patience use bicubic, deinterlace their video, and use CVDs & SVCDs over VCDs.

If shooting for quality why distroy up to half the image by de-interlace? The loss of both spatial and temporal image is not a good idea expecially since you are aspitring to use formats that fully support fild based encoding.

[bbb]

Please, don't mistake erroneous attempts at deinterlacing with proper deinterlacing. Open your mind and learn something new by giving serious effort in trying something new/difficult.

By taking the time to learn proper inverse telecining and deinterlacing, one can create cleaner video clips that play great on computers as well as on standalones.

Unfortunately, too many are too lazy to learn how to do so and then use the sour grapes excuse not to do so. Also, these lazy ones are stuck with playing their video clips on standalones to look good. This is really limiting. And tragically unnecessary.

I've tested progressive encoding and interlaced encoding with many AVI caps, there is no visual degradation in video with progressive encoding with proper deinterlacing. Progressive encoding after a proper deinterlacing creates a far cleaner and more versatile video stream without degradation in quality, smoothness, and clarity.

Moreover with additional proper filtering, all my clips are 100% free from interlaced artifacts.

[energy80s]

It must be great being YOU

[snowmoon]

Oh, IVTC...

That's different. Restoring the original progressive stream is diffrent than destroying a video stream through de-interlace. But that only works for movies and the occasional TV show. Most TV shows have horrible patterens that make them poor candidites for IVTC.

If a show is interlaced how is reducing it's rate from 60 fields/second to 30 frames per second going to improve it?

Outside of exclusive use on a PC there is no reason to de-interlace interlaced materal ( not counting telecine ).

[offline]

Outside of exclusive use on a PC there is no reason to de-interlace interlaced materal ( not counting telecine ).

I disagree. The future is progressive. Why lose picture
quality on your future progressive scan TV because you
were in too much of a hurry not to deinterlace. Are
you going to rely on hardware to deinterlace your
favorite TV shows? If you do it correctly you get
a superor picture - i've tested this myself with
commercial quality sources.

[bbb]

offline, hits it right on the mark. The future of digital televisions is coming soon. Those who encode for their analog televisions will be SOL when they videos look horrendous with interlaced lines on digital sets.

snowmoon, doesn't seemed to have tried proper deinterlacing (for film and TV shows) on AVI caps. His ongoing statements that deinterlacing causes destruction is without merit. If he has bothered to learn proper deinterlacing and compared an interlaced to progressive encoding, he will realize that he would have to eat his words.

I have produced both types and time and time again my deinterlaced progressive videos beats the interlaced encoding on the computer monitors (zero interlaced artifacts/lines show up). On analog sets, both are identical (no visible loss of quality).

It may take longer but why bother encoding in a soon to be outdated, inherently flawed video format (interlaced video)? Laziness to learn and impatience are poor excuses.

[snowmoon]

offline, hits it right on the mark. The future of digital televisions is coming soon. Those who encode for their analog televisions will be SOL when they videos look horrendous with interlaced lines on digital sets.

Ummm... the most popular HDTV format is..... 1080 interlaced. Oh my god what the hell are they doing.. I guess that foramt would look like crap on my 42" HDTV widescreen.

snowmoon, doesn't seemed to have tried proper deinterlacing (for film and TV shows) on AVI caps. His ongoing statements that deinterlacing causes destruction is without merit. If he has bothered to learn proper deinterlacing and compared an interlaced to progressive encoding, he will realize that he would have to eat his words.

I did for quite a while, but it DID affect quality and my encodes have been better since I got over my faciniation with doing everything progressive. It's a matter of getting the right setting on the encoder.

I have produced both types and time and time again my deinterlaced progressive videos beats the interlaced encoding on the computer monitors (zero interlaced artifacts/lines show up). On analog sets, both are identical (no visible loss of quality).

It may take longer but why bother encoding in a soon to be outdated, inherently flawed video format (interlaced video)? Laziness to learn and impatience are poor excuses.

Interlaced is not outdated, and to my knolwdge all progressive sets handle interlaced video great. Without a doubt progressive set are more than likley to have the better progressive upscaling circuits.

Lazyness is not a part of the equation. I have spent the last three days trying to get one encode to come out right. I'll keep going till it does look right. And I would not start pointing fingers about ignorace.

What size TV do you own? I have a 80" projector in the basement and a 42" widescreen HDTV set and it shows all the flaws. Both handle interlaced materal better than most de-interlaced materal. I am still not refering to telecined materal whose native format is progressive and it is possible to re-capture that with a good IVTC.