BANDING..still..

[zanaitoryoushi]

I have another question on banding. I have been encoding movies for my psp and it seems this and aliasing are my major problems.

I would care less for aliasing for now since banding is much more irritating. Aliasing would often appear on areas where the edges are heavy in color (e.g. red, green, etc.) so i can forgive that. Anyway..

I've tried doing dithering/debanding on the videos i encode and the results are good. But i'm very sensitive to the grain that it adds. I wish they were never there, lol. Well that's how it get rids of the banding anyway, or rather hides it but if i can find a better solution then that would be great.

I've read on wikipedia that raising the number of bits per color channel will help on eliminating the banding but i'm not really sure how to do that? does it mean that i need to raise the bitrate? i hope someone can interpret that for me. My output file size is above 1.5 GB always and i'm currently using this: removegrain, sharpen(.7), Flimslevels (i think that's what it is called) - set to 2.5, tweak(with sat=1.117,bright=.7 & con=1.117), dfttest then addgrain+the default debanding tool of ripbot with a blur of 0.01.

Given the source is already good, my output is also good in comparison save for the grain. I want to remove it and fix the banding problem with a different solution. I don't really care about the output size.

Also does it depend on your source? The video i'm converting was encoded at 1509 bitrate. I'm using CQ 16 in ripbot and the banding still shows. Does switching mode to 2-pass and using the same bitrate of 1509 would prove better? Sorry i just had a thought of it now and unfortunately i'm at work so i can't test it or even provide a clip of the video or a screenshot. It really bugs my mind so i decided to ask here. Anyway, thank you in advance.

[poisondeathray]

I would care less for aliasing for now since banding is much more irritating. Aliasing would often appear on areas where the edges are heavy in color (e.g. red, green, etc.) so i can forgive that. Anyway..

This actually might be the chroma upsampling by the PSP, not the video .

Post a sample

I've tried doing dithering/debanding on the videos i encode and the results are good. But i'm very sensitive to the grain that it adds. I wish they were never there, lol. Well that's how it get rids of the banding anyway, or rather hides it but if i can find a better solution then that would be great.


I've read on wikipedia that raising the number of bits per color channel will help on eliminating the banding but i'm not really sure how to do that? does it mean that i need to raise the bitrate?

"Banding" can be from low bit depth or compression artifacts . These are different things. When most people see "banding" it's usually from compression artifacts.

8bit means you can have a range of expression from 0-255 or 256 values 2^8 (eg. black to white can be 0-255) . 10but means 2^10 or 0-1023 or 1024 values. You can thing of it as having smaller gradations or "shades" between the extremes. Because there are more "shades" and there finer "steps" and range of expression, the gradient is smoother.

It doesn't necessarily help if you already have banding or compression artifacts in the source. Converting an 8bit source to 10bit doesn't mean it suddenly mean it's a "real" 10bit source. An analogy would be if you upscale a SD to HD source, the HD upscale will still be SD content. Higher bit depth helps more when you start with higher bit depth

PSP doesn't support higher bit depth (8 bit only ) anyways. You can use 10bit AVC, but no devices play it back (computer only). 10bit AVC does help with banding (even from an 8bit source), and you do get higher quality at a given bitrate . The main reason is internal precision for calculations, not range of expression 10bit values (0-1023)

Ordered dithering instead of adding grain randomly (using add grain like you are now, or gradfun2dbmod) will usually be more pleasant (looks less like noise). Also bitrate requirements will be lower since it's more of a fixed pattern

My output file size is above 1.5 GB always and i'm currently using this: removegrain, sharpen(.7), Flimslevels (i think that's what it is called) - set to 2.5, tweak(with sat=1.117,bright=.7 & con=1.117), dfttest then addgrain+the default debanding tool of ripbot with a blur of 0.01.

Given the source is already good, my output is also good in comparison save for the grain. I want to remove it and fix the banding problem with a different solution. I don't really care about the output size.

It can't be a good source if you're using dfttest or all those filters. A "good" source wouldn't need any filters at all.....

Also does it depend on your source? The video i'm converting was encoded at 1509 bitrate. I'm using CQ 16 in ripbot and the banding still shows. Does switching mode to 2-pass and using the same bitrate of 1509 would prove better? Sorry i just had a thought of it now and unfortunately i'm at work so i can't test it or even provide a clip of the video or a screenshot. It really bugs my mind so i decided to ask here. Anyway, thank you in advance.

It always depends on the source.

Higher bitrate always helps

Some settings like AQ may help as well (higher values will reallocate to flat areas that are prone to banding)

[jagabo]

I've tried doing dithering/debanding on the videos i encode and the results are good. But i'm very sensitive to the grain that it adds.

8 bits (per channel) of RGB, and even less of YUV, are not enough to make visually smooth gradients. You can see this for yourself by creating smooth gradients using any 8 bit paint program. Noise is required to break up the bands. You're obviously removing too much grain from your sources.

I've read on wikipedia that raising the number of bits per color channel will help on eliminating the banding but i'm not really sure how to do that?

No it means you have to use software that works in more than 8 bits (ie, 10 bit, 12 bit, etc.). People have good luck with x264 in 10 bit mode. But if you use that you are restricted to PC playback as most standalone hardware can't handle playback of 10 bit h.264 at this time.

[zanaitoryoushi]

@poisondeathray and jagabo

thank you for the replies (still doing a couple of things though), i've read through your explanations/suggestions and it really is useful..
about the filters, i added removegrain for speed and dfftest for a little bit of compression, but maybe i'll remove those, probably removegrain and dfftest removes too much grain and noise thus the banding appears (the source i'm getting are blu-ray rips ranging from 2-4 GB so i guess i'll remove those other filters and keep the other functions i'm using for the visual adjustment), also i'll try converting it straight up with just changing the bitrate..probably it'll remove the banding problem, so depends on the source.. converting it with a lower bitrate than the original will obviously cause banding to appear..and yeah i've read that 10bit somewhere but like what you guys said, it won't play on the psp..anyway, thanks a lot for the replies, i really appreciate it. Sorry if i can't comment through everything that you guys said, i'm currently working, but thanks for everything..this clears up some things, i guess i'll progress better with my encoding...many thanks!


also i'm not sure about the AQ thing, this is the first time i heard it, but thanks guys, i am removing a lot of noise with the filters that i am using, i'll try without them the next time i encode..thanks again.

[jagabo]

Yes, using --aq-strength from 1.0 (default) to ~2.0 will increase the amount of grain retained in the x264 compression. But it will also increase the bitrate requirement.

[zanaitoryoushi]

Yes, using --aq-strength from 1.0 (default) to ~2.0 will increase the amount of grain retained in the x264 compression. But it will also increase the bitrate requirement.

oh right, that's what AQ is. Silly me, the word was all in upper case and suddenly i don't know it, lol
yep, i personally use 1.2 as my aq-strength. So that plus the omission of the grain/noise removal plugins = lesser banding, i will def do that next time given that i'm not a whole lot busy with work. =)

[zanaitoryoushi]

one last question..my source uploads a copy of a blu ray movie in two different types..a 1080p and 480p version..

1080p versions are encoded with 3509 bitrate and 1509 for the 480p version

my question is (i know the answer it's pretty obvious, but anyway) should i pick 480p versions over 1080p? i mean for a psp movie, i can basically encode the movie with the same bitrate and it may not exceed 2.5 gb, however, i don't think the same will be true if i get the 1080p version, plus, if i'm right, even with the fancy 1080p label, since i'm converting the movie for a psp and i'm still aiming for a size not exceeding 3 gb..i think lesser bitrate encoding for that version is more prone to banding and other sorts of unusual problem right?

(btw, i'll be testing this this Saturday, kind of busy this week in the office)

I appreciate the help a lot guys! thanks in advance..

[poisondeathray]

It's hard to say with just those numbers.

And it will depend on the source, how it was encoded, how it was processed. There are too many variables . Try it and find out

Mathwise, the 480p version should have more proporational bits per framesize. But this doesn't necessarily indicate anything.

In the grand scheme, you 're talking about psp here, so it's not going to make a big difference


The only certain thing is both will already have artifacts compared to the original blu-ray

[zanaitoryoushi]

And it will depend on the source, how it was encoded, how it was processed. There are too many variables . Try it and find out

Precisely what i will do, i have both copy of the movie i want already, looking forward to my encoding on Saturday
I have converted the 480p already with the setting i indicated above, i'll try both next with the omission of removegrain and dfftest converted in 2-pass with high bitrates. And your right, there are too many variables, maybe the 1080p would prove to be better. i'll just try. Thank God for mkvmerge, else i can't really test them both and get quick (actual) results.