100fps: How to deinterlace Video the right way. Examples, Tricks, Comparisons

Which video compression codec is better?
Why comparing still pictures doesn't help much if you don't know what to look for.

Important: I do not compare any video codecs here. I give you just a few thoughts ABOUT codec comparison. Use this information I give you BEFORE you read/make video codec comparisons.

OK, then let's go:
Please look at the following pictures and tell me which one you like the most (look at the face):




Look at this picture. Pretty sharp and with lots of details. Compare the leaves in the background with the leaves in the next picture. Compare the balloon stars and compare the face. Compare the clothes. Everywhere you see more details.
But maybe you like this picture more?
How about this face? It shows a soft skin. If you look carefully over the face you will notice, that no detail is missing, they are just smooth.

 

Now, to tell you the truth, the picture in the middle is the original. The picture to the left shows LESS details than the original one. It simply sharpens the edges, by turning light gray to white and dark gray to black simply speaking. So you LOSE details by losing black-white gradients. That's what they do in cooking books (to make the food look sharper and more crisp). You can see that white border of the girl's cap (her name is Christina). There isn't anything like that in the original picture. This shining around dark areas is called "Aura".

The picture to the right is blurred to your satisfaction. That is exactly what fashion magazines are doing with faces and what they do in movies, TV shows and TV sets to fight snow and face wrinkles. Jennifer Lopez' (a former background dancer for Janet Jackson) face in this scene (and the whole video) is really smooth:

 

The following face is so smoooooth, no boy-friend will ever stick to her:

(This pic had the name Chloe Jones at the upper right corner, so I guess that is her name. What a lonely girl...)
This is an interesting picture, however, because they made her skin smooth, but sharpened her eyes, lips and hair.

   

So if you see a comparison of codecs and one still seems to show more details than the other (like the left picture here), it does not mean, that it a better codec. It could simply sharpen (and thus lose details). And if you see one codec that shows less blocks than the others then it could simply mean, that it blurs (like the picture right).

The best way to fight codec details loss is to set the parameters to "Constant quality" instead of "Constant bitrate".

   

There is another thing about codec comparisons. They are not fair:

DivX is an MPEG codec. That means that it shows a certain kind of compression artifacts:
Typical MPEG artefacts are blocks and mosquitoes. Here are 2 captures from the music clip George Michael & Whitney Houston "If I told you that"


MPEG produces Blocking Artefacts in scenes with many changes (=fast movements, scene change, a lot of motion). See the capture above (Britney Spears "I love rock 'n' roll"): This is a frame where there was a scene change, there's a lot of funny paper flying around and she's dancing.


Really blocky. Would you believe when I told you, that the sound to this picture was aired in mono (instead of stereo)? 64kbit/s? Mpeg Layer 2? On a music channel (MTV India)? Modern TV Quality.


As another artifact MPEG produces Mosquitoes at sharp contrasts like mouth/skin, ocean/sky, dark hair/bright wall. See around Whitney's mouth and eyes (no, it's not because she just came from a cocaine session).

Sometimes you see compressed text in the internet like these message boxes with their typical artefacts. I created these sample message boxes:


JPEG artefacts around the text.


GIF artefacts: Clear text, but less colors (see the title bar and the question mark).

 

If you take footage, that has already been compressed by another MPEG codec, then you amplify those artefacts.

So if you record a digital TV broadcast and then encode it with DivX this means, MPEG-2 (broadcast) is encoded with MPEG-4 (DivX). The same applies to DV (which also uses MPEGlike compression methods) and DVD (which is MPEG-2). VCD is MPEG-1 and SVCD is MPEG-2.

But not only that compression artefacts are increased, to compress them it takes more bandwidth (in theory), of course: Because how should the codec know what is an artifact and what is part of the image? So it has to encode the artifacts also. Thus you increase artefacts AND file size. I have made experiments though, that show, that file size actually decreases when recompressing. Why? Because DivX blurs and darkens footage when compressing and compresses better when blurred. So each recompression makes your movie more blurred (and darker) and thus compresses better.

That means, to be fair you'd have to compress a RAW clip and compare then.

To see how re-encoding (re-compressing) affects the quality see my recompression comparison.

   

There is another thing about codec comparisons. Quality vs. Filesize:

One can scrutinize for quality at high bitrates or for file size at low bit rates. Please always pay attention what the purpose of the test is: "How good does the picture look when we compress the content to death?" or "What is the bitrate I need to make the content look DVD-like". These are 2 different approaches.

   

100fps: How to deinterlace Video the right way. Examples, Tricks, Comparisons