newbie question about "loss"

[chuckcintron]

Love HandBrake!

In the past I have just been ripping my DVDs to VOB files then my little media player "plays" the folder. These files from a typical DVD are usually around 6GB in size.

I experimented with HandBrake, encoding with h.264 and using the "high quality" preset and a quality level of "18", and putting the result into a MP4 container. Resulting file size is around 1.5GB in size.

The resulting video looks fantastic on my 55" plasma TV as well as on my 120" LG PF1500 projector.

Here is what I don't understand: 75% of the original data is "gone", file size going from 6GB down to 1.5GB. So, what exactly do you "lose" in this process? Is my brain being tricked into thinking the resulting video looks just as good as the original? Or is the transcoding just so excellent that it can maintain par equivalency to the original content?

I realize the full answer to this is probably a doctoral thesis in computer science but generally/simply speaking -- I'd like to understand what "goes away" during this process?

[mduell]

Part of it is indeed tricking your brain. x264 has a good pyschovisual model to allocate bits where you'll notice them.

Also newer more efficient and more computationally demanding encoding methods. The video format for DVDs is 20 years old.

[Deleted User 13735]

Handbrake puts the bits where they are needed, quite simply.
Your DVD s use mpeg 2 compression, which is many times less efficient than x264, meaning it puts a lot of bits where they do no good.
The Handbrake Wiki is the place to start your learning curve. Welcome to the forums.

[chuckcintron]

got it, thanks. I like having my bits where they belong

Will read up at the wiki - good stuff.

[Devore]

A good analogy is CD vs MP3. The MP3 is much smaller, but still sounds almost as good, for most use cases. In the audio world, we have psychoacoustics, which is how we can throw away information, but still fool the brain into thinking nothing has been taken away.

In the video world, it is the same. There are tricks we know that we can use to foll our eyes and brain. This involves throwing away some detail, using more efficient compression techniques (mpeg2 vs h264 vs h265, etc) or reducing bitrate/depth/resolution. For example, DVDs use a fixed bitrate, even for parts where very few bits are needed, that's a lot of waste.

[Deleted User 13735]

DVD movies can be either CBR or VBR, the latter still having far less efficient compression than h264.

[Hand On Tha Brizzake]

It also depends on the specific parameters of the profile you use. I customised mine so nothing is removed and compression is most efficient. H.265 [x265] is my codec of choice and the immense hours my system requires to transcode the DVDs' feature titles is worthwhile to me. I have never used h.264 [x264] thus I know nothing with regards to comparison to the former using my exact settings. If your system will play h.265 [x265] I wholeheartedly recommend you re-transcode from the source if possible. Alas, if you wish to know my specific settings that maximise efficiency and minimise quality loss I will be glad to assist if they will do so.

[Deleted User 13735]

I respect your point of view, but very high quality x264 encodes from DVD movies are already small enough, usually <=1GB.
That said, the exponentially greater render times needed to get the same quality with more efficient x265 compression is simply not worth it to me.

[Hand On Tha Brizzake]

I never used the codec thus I believe you. With regards to your MP3 analogy above, I can clearly hear the difference between MP3 with a sample rate of 48000 hertz and bit rate [variable or constant] of 320 kilobytes per second and any lossless file of the same sound. If I could see there is no telling how the video side of things would parallel.

[Deleted User 13735]

With regards to your MP3 analogy above, I can clearly hear the difference

What???
Just what are you smoking (it's not working!)???

[Hand On Tha Brizzake]

My mistake. I just realised another person made that analogy.

[Deleted User 13735]

Come up for air. It's just fine to sing the praises of x265, but right now it's about file size vs. encoding time. Nothing else.
Optimally compressed x264 from DVD or Bluray is the same quality as optimally compressed x265 from the same source. Optimal compression ratios will be different, but in both cases, are >RF=0.

[Hand On Tha Brizzake]

Point well taken. Codec notwithstanding, would you agree compression with regards to file size is most optimal when set to Placebo? Time required for the system to encode is not exactly optimal but I guess it depends on the system used. A twelve core late 2013 Mac Pro would certainly require less time than a Core2 Duo late 2009 iMac.

[Deleted User 13735]

Codec notwithstanding, would you agree compression with regards to file size is most optimal when set to Placebo?

Errm, compression and file size are direct references to the same thing.
Y = -m(X)

And no, I would not agree with that in 99.999% of cases.
Again, DVD or BluRay compressed source (8 bit 4:2:0 Y'(CbCr)) simply does not warrant it.
A nontrivial advantage would be hard to establish, I suspect, but you're welcome to try if you are convinced.
Placebo is the 'kitchen sink" setting. Put there to placate the compression-obsessed (if trhe name isn't enough of a tip-off).
I'll run my tests again when we're all working with 4K 10 or 12 bit 4:2:2 or 4:4:4 source. Until then . . .

[mkrjf]

OK so I have 4K 10 bit 4:4:4:4 input. I also have a SUHD tv capable of displaying 10 bit 4:2:2 or 8 bit 4:4:4 as well as 10 bit 4:2:0.
So it seems most of the ffmpeg / x26* packages compile to 8 bit or 10 bit processing. Nothing can be switched in software.
Is that true for handbrake as well? If so the next generation of high quality content will be 10 bit 4:2:0 at least (UHD BD standard recently finalized and players due this winter).
Anyone who can afford Panasonic GH4, especially with shogun will be posting 4K content to youtube let alone UHD BD. Actually they already are!

So when would handbrake be able to support 10 bit video? Other than performance it has been shown (see Ateme white papers) that best results are using 10 bit code (even for 8 bit input). Just need options to convert final stage to 8 bit and algorithms for chroma sub-sampling.
Seems none of the ffmpeg or similar does bit depth conversion or sub-sampling conversion - with possible exception of prores input.

Any thoughts, comments?

[Deleted User 13735]

Any thoughts, comments?

Yes, you're still about two years out, my friend.