How to reduce blockiness on soft backgrounds

[^JazzMan^]

I was wondering if anyone could suggest any possible ways to "eliminate" the appearing of blocks on soft backgrounds, if the avg bitrate is arroound 3000kbps?

Thanks forwardly

cheers

[Digga]

I was wondering if anyone could suggest any possible ways to "eliminate" the appearing of blocks on soft backgrounds, if the avg bitrate is arroound 3000kbps?first of we need to know what encoder you're useing ;)

e.g. for CCE:
Quality_prec
Also known as Image quality Priority in CCE 2.5 and Quantizer characteristics in CCE 2.66 +. This setting ranges from 0-64, but is scaled for CCE 2.5 which uses a 0-100 scale.
This setting controls whether CCE gives priority to the fine details of the image, or evenly colored flat areas. A low setting will give priority to details, but could result in blockiness or color banding. A high setting will favor flat areas, but could result in edge artifacts.

also, CPC has a reputation to soften the image more than e.g. CCE i.e. don't cause so much blockiness.

[^JazzMan^]

So pushing the Quality perc to higher values would help reducing the blockiness? What about using HC Encoder?

[ux-3]

What setting are you using, when avg. bitrate is in the 3000-3500 region?

[^JazzMan^]

When encoding with HC Encoder in 3000-3500 kbps range, I usualy use QLB matrix, VBR_Bias set to 30

[rahzel]

When encoding with HC Encoder in 3000-3500 kbps range, I usualy use QLB matrix, VBR_Bias set to 30
3000-3500 is a pretty low bitrate range imo and a vbr bias of 30 is a little too high IMO. use the default 20, or maybe 25.

also, try the AUTOQ2 matrix. i've already imported it with RME (rebuilder's matrix editor) so you can just download the file ive attached to this message, unzip it and put it in your "Matrix" folder, in your DVD-RB directory.

i really like AUTOQ2 with CCE and HC.

[l8nights]

I'll have to give that a try Rahzel (when it's approved) I've heard you preaching enough, now my interest are peaked ;)

any special circumstances it excels in??
I ask because I have some older titles still waiting to be encoded ie a couple from teh stanley kubricks collection
L8

[rahzel]

I'll have to give that a try Rahzel (when it's approved) I've heard you preaching enough, now my interest are peaked ;)

any special circumstances it excels in??
I ask because I have some older titles still waiting to be encoded ie a couple from teh stanley kubricks collection
L8
i use it in all my projects. its great for removing blocking and noise, but it doesnt seem to remove much detail.

i did the movie gladiator with HC and CCE at about 3380 bitrate, and with HC, it was really blocky, while CCE was really blocky and noisy. i used the AutoQ2 matrix, without any filters, and it looks great.

i believe the creator of AutoQ2 (amnon82) said that AutoQ2 works best with bitrates between 1800-3500, but i just use it for all the matrix settings in DVD-RB (main movie, low bitrate, very lowbitrate etc).

im kinda new at DVD-RB, so if anyone who tries AutoQ2 and can recommend me a better one, by all means do so.

[manono]

Hi-

Is this the AutoQ2 matrix:

08 16 19 22 26 28 32 38
16 16 22 24 28 32 38 44
19 22 26 28 32 38 44 48
22 22 26 32 38 44 48 54
22 26 32 38 44 48 54 64
26 32 38 44 48 54 64 74
32 38 44 48 54 64 74 84
38 44 48 54 64 74 84 94
16 20 24 28 36 42 46 52
20 24 28 36 42 46 52 58
24 28 36 42 46 52 58 62
28 36 42 46 52 58 62 68
36 42 46 52 58 62 68 78
42 46 52 58 62 68 78 88
46 52 58 62 68 78 88 99
52 58 62 68 78 88 99 99

If so, you don't have to wait for the attachment to be approved. I had it in a matrix collection listed as AUTO-Q2. Looks like a pretty low bitrate matrix to me. It has some similarities with the Standard matrix, but retains even less detail. Should be good for low bitrate extras, and maybe long or hard to compress movies, but I don't think I'd want to use it for a high quality movie backup. Just my 2 cents. :)

[rahzel]

yeah, thats the AutoQ2 matrix.

well, all the movies i use DVD-RB with will be pretty low bitrate (around 3000-low 4000's).

even if the matrix indicates it removes detail, i have done many tests by taking SEVERAL screenshots and zooming them in 400%, and its VERY close to the original, better than most of the matrices ive used. CCE SP's default matrix was very margionally better than autoq2 in high bitrate scenes, though. i have not done a HIGH bitrate project yet, so i cannot compare it to a higher bitrate matrix, like FOX entertainment.

how do you tell if a matrix is good for low/high bitrates etc from reading the numbers? do lower numbers=for higher bitrates? what matrix would you recommend for bitrates between 3000-4000 and for noisy and/or non-noisy movies.

[dialysis1]

If I'm not mistaken, it's the same as AVAMAT7 matrix.

[rahzel]

edit: @ dialysis1, yes, autoq2 is aka AVAMAT7.

sorry for getting off topic, but i did batman begins with DVD-RB Pro, HC 0.18 (BEST profile), the AutoQ2 matrix for all settings, decoder idct xvid simple mmx, everything else default. it was about 69% with about 4000 overall bitrate i think.

here are some comparisons... tell me what you think.
(source=top, hc=bottom)

http://i8.photobucket.com/albums/a31/rahzel54/s.jpg
http://i8.photobucket.com/albums/a31/rahzel54/hc.jpg

http://i8.photobucket.com/albums/a31/rahzel54/s2.jpg
http://i8.photobucket.com/albums/a31/rahzel54/hc2.jpg

http://i8.photobucket.com/albums/a31/rahzel54/s4.jpg
http://i8.photobucket.com/albums/a31/rahzel54/hc4.jpg

http://i8.photobucket.com/albums/a31/rahzel54/s6.jpg
http://i8.photobucket.com/albums/a31/rahzel54/hc6.jpg

[techmule]

Or you can try AutoQmat (http://autoqmatenc.com) Encoder, for extremely low bitrates (i.e. avg below 2000Kbps) enable its QmatOp. Download the latest Beta release only.

With QmatOp enabled you dont have to worry about selecting the right matrix, thats what QmatOp does and that too tailored for each segment of the movie.

low BR = autoqmatenc= Good results :)

[l8nights]

If so, you don't have to wait for the attachment to be approved. I had it in a matrix collection listed as AUTO-Q2. Looks like a pretty low bitrate matrix to me. It has some similarities with the Standard matrix, but retains even less detail. Should be good for low bitrate extras, and maybe long or hard to compress movies, but I don't think I'd want to use it for a high quality movie backup. Just my 2 cents. :)

hmmm I didn't have it til now thnx manono!

now you wouldn't reccomend this except for low BR?

I'm actually looking for a matrix to use for animation, I didn't like the standard one's, I've been using CPC for preffered results with animated CG titles, has anyone any suggestions??

[Voodoochild]

Speaking of Matrix, I was wondering what the quantization matrix in DVD-RB mean in the advance options there are
main feature matrix
low bitrate
very low bitrate

Is that mean the CCE will use different Matrix (if I choose one ofcorse) for each segmant according to it's bitrate??
meaning, If I have segmant in 2800 bitrate and another in 4200 bitrate , CCE will use the main for the 4200 segmant and the low for the 2800 segmant ??

10x in advance Elad.

[jdobbs]

Yes. The selected matrices will be used based upon segment bitrates -- assuming your encoder will accept alternate matrices (CCE SP, HC, QUENC, and AutoQMatEnc will all accept alternate matrices). AutoQMatEnc and CCE also have "Automatic" modes in which the encoder will select a matrix.

Note: The CCE automatic mode is accompanied by a warning that its use may not be compatible with all DVD players -- but I haven't heard any reports of incompatibility as yet.

[Voodoochild]

Note: The CCE automatic mode is accompanied by a warning that its use may not be compatible with all DVD players -- but I haven't heard any reports of incompatibility as yet

for now I want use the hidden option... maybi later on, when I'll have more expeirience with matrix.
I'll try to encode an old westren I have "Rio Bravo" with bitrate between 4000-5300 with fox matrix and see how it goes, 10x very much
Elad.

[lambo]

@rahzel

ecxuse me... could you tell me wich program are you using to take snapshots of the voriginal and compressed movie in the same exact frame? virtualdubmod?

thanx

[rahzel]

i believe you can use vdub, but i used media player classic, and just paused at the corresponding times.

[bennynihon]

i believe you can use vdub, but i used media player classic, and just paused at the corresponding times.

rahzel, the screens look good! Any chance you can freeze and capture a more detailed image, such as the closeup of someone's face, or something similar? I would like to see how much detail the autoq2 matrix removes from scenes that contain more high frequency, fine details. Thanks!

[rahzel]

rahzel, the screens look good! Any chance you can freeze and capture a more detailed image, such as the closeup of someone's face, or something similar? I would like to see how much detail the autoq2 matrix removes from scenes that contain more high frequency, fine details. Thanks!
you could just save the screenshots to your harddrive (right click, save as), open with MS paint, and zoom in. this is what i do to compare shots.

screenshots arent a very good way to compare, but i myself have a hard time looking for a difference when the video is playing.

[^JazzMan^]

i believe you can use vdub, but i used media player classic, and just paused at the corresponding times.

What if the encoded movie has different GOP structure than the source, so when you freeze the motion on certain timing you can get, for instance P frame from the source and B frame from the encode or reverse ; AFAIK P frame captures always look better than B frames. I think it would be better to do the capturing with virtualdub, that way you can see which frame you capturing, though IMHO following the motion is the best way to make the comparison.

cheers

[jdobbs]

Looking at frames is never a good way to compare video. It's like trying to study a waterfall by looking at a drop of water.

[JFerguson]

Hmmm... Yeah, I didn't need stills for King Kong 2005. Maybe I should try this AutoQ2 thing? CCE and ProCoder at standard settings didn't do too much for me...

[Voodoochild]

Hmmm... Yeah, I didn't need stills for King Kong 2005. Maybe I should try this AutoQ2 thing? CCE and ProCoder at standard settings didn't do too much for me...

Hmmm... Yeah, I didn't need stills for King Kong 2005. Maybe I should try this AutoQ2 thing? CCE and ProCoder at standard settings didn't do too much for me...

Hi JFerguson, AutoQ2 AKA Avamat7 is better for high compression, it will probabaly remove more blocks and noise from your dvd since it compress more, but retain bit less details. if you look at the Matrix, It has high values meaning the treshold is high and in some sence it will compress more, causing maybi softer image and less bolckiness. How ever if you can and compression is not to high, try other Matrix that has lower tresholds and preserve bit more details. Also if you use RME to compare Matrix keep in mind that the bitrate RME encode the segmant are different then when it's encoded with DVD-RB, since the segmant probably need some compression making CCE use lower bitrate then the original.

[JFerguson]

@Voodoochild:

Yeah, I might try that. I did Dune R1 Extended Edition the other day with 4-pass CCE and standard settings and it came out pretty good. DISC 1 with extras was 3229kbps on the main movie, DISC 2 was 2943kbps. These were slightly noisy to begin with -- I guess it all depends on the source...

[Voodoochild]

I guess it all depends on the source...

for you, you can try use filters in avisynth editor. for my self if it's too noisy I use hqdn3d(2), it removes noise without loosing much details. If I do that I use 2 passes instead of 3, cause it takes more time...... :-)

[bennynihon]

Looking at frames is never a good way to compare video. It's like trying to study a waterfall by looking at a drop of water.

I'm not in agreement with this. While it's true that looking at still frames doesn't accurately reflect how well it will look as a video, you can much more easily discern difference in detail by looking at still frames. Certainly after watching it in motion, with the filtering effects of your display and your own visual system, that subtle difference in detail that you can pickup by still frames may no longer be appreciable. In which case you could argue "what's the point of comparing still frames, if you can't even notice a difference in video?". But as videophiles we get comfort in just knowing (even if we can't perceive it in motion) that one video did a better job in preserving the detail of the original than another.

I agree that there's more to video than just a series of still frames. It's entirely possible that one video looks great at each still frame, but somehow botches the fluidity of frame to frame (by perhaps messing up the motion estimation). But since we're comparing outputs from the same encoder, but with different filters, it actually is a good way to compare video by looking at still frames to see the differences in detail preservation.

A better analogy, would be one taken from audio. You could say that it's silly to compare two audio streams by viewing their analog waveforms to the original. It may be entirely possible that the two encoded audio streams sound virtually identical. However, by comparing the waveforms of each encoded stream to the original, you can see just how well each stream comes to accurately recreating the original.

[ux-3]

A better analogy, would be one taken from audio. You could say that it's silly to compare two audio streams by viewing their analog waveforms to the original. It may be entirely possible that the two encoded audio streams sound virtually identical. However, by comparing the waveforms of each encoded stream to the original, you can see just how well each stream comes to accurately recreating the original.

You chose a good analogy, but unfortunately it works just the opposite. You could have an audio wave form that was 99.9 % accurate yet would drive everyone listening nuts. All you need for that is a little clipping. And you can have a waveform, that has absolutely no resemblance to the original wave form, yet it would sound allmost identical to the human ear. The later is the case in high bitrate mp3. Due to the loss of some phase relations, the resulting wave looks completely different in time space, but is virtually unchanged in the frequency domain. The former wave form would hardly deviate from the original in the time domain, but show significant deviation in the frequenccy domain. And the human ears sensor array works in the frequency domain.

Unless you plan to view your movies as a slide show, two very important effects are completly ignored in a frame study.
1. Thinks may look as if they are viewed through hot air.
2. Large unicolored surfaces could pulsate or change the tone borders very noticably.

Both are common problems and both can't be seen in single frames

greetings
ux-3

[rahzel]

if you compared several screenshots (lets say, 7-10) of 2 different matrices and all 7-10 looked better with 1 matrix, would that be a good enough indication that the matrix is better than the other?

[Voodoochild]

if you compared several screenshots (lets say, 7-10) of 2 different matrices and all 7-10 looked better with 1 matrix, would that be a good enough indication that the matrix is better than the other?

I do both, I'm taking 10 pics and compare them, and I also look at the video, looking for blocks, noise , flowless transitions between sences, colors etc'.....

usually what you see on the computer screen, you want see at the TV, but as you said it's good to know that you made the Maximum to preserve details and get the Highest quality/compression you can.

[rahzel]

the thing im confused about is i've done about 5 movies now, and took probably over 30-40 screenshots (for all 5 movies) and all of them looked best with AVAMAT7, compared to all the other matrices ive tested. when i look at all my screenshots, i zoom them in 200x/400x to see the difference, as its impossible to see a difference when its not zoomed in.

i know comparing a single screenshot isnt a good way to compare, but surely if AVAMAT7 appears to be better after 30-40+ screenshots, it has to be better, doesnt it?

btw, in the 5 movies i've tested, the overall bitrates were roughly 3385, 3459, 4600, 4200, 3000, so the bitrates were varied with all the movies i tested with. i've tried FOX Entertainment all the way down to MPEG standard, CCE with MPEG Standard and AQM enabled, QLB, and this is for both HC and CCE btw.

is there anyway that the numbers in a matrix could be mis-leading? AVAMAT7 appears to be for low bitrates, but it seems to be closer to the source than any other matrix i've used. i also noticed that AVAMAT7 is similar to MPEG standard at the top half of the matrix table, but has much higher numbers at the bottom of the matrix table; does that mean it will remove more detail in certain situations? if so, what situations?

[Voodoochild]

btw, in the 5 movies i've tested, the overall bitrates were roughly 3385, 3459, 4600, 4200, 3000, so the bitrates were varied with all the movies i tested with.

1. do you remember what was the compression for each movie? because the finall result is very much depend on compression amount, higher bitrate matrixes not so good with high compression.

2. did you test screenshots after dvd-rb encoding or RME encoding, they do not use the same bitrate, there for final result is different.

3. I have also tested full dvd backup with RB once with cceaqm on, default, and avamat 7. In my case 2 movies "Rio Bravo", "The last Samurai" bitrate were 3900 (last samurai), 4000-5000 rio bravo. compression were 70-73%.
cceaqm=1 gave me the best results. avamat7 did not removed more blocking then the others,
both cases the output chosen matrix by cceaqm were very close to fox entertaiment.

so I guess it's more of each one preference and what looks better to his eye, i think ...

I'm wondering how cceaqm will function in high compression though ?

[rahzel]

1. do you remember what was the compression for each movie? because the finall result is very much depend on compression amount, higher bitrate matrixes not so good with high compression.

2. did you test screenshots after dvd-rb encoding or RME encoding, they do not use the same bitrate, there for final result is different.

3. I have also tested full dvd backup with RB once with cceaqm on, default, and avamat 7. In my case 2 movies "Rio Bravo", "The last Samurai" bitrate were 3900 (last samurai), 4000-5000 rio bravo. compression were 70-73%.
cceaqm=1 gave me the best results. avamat7 did not removed more blocking then the others,
both cases the output chosen matrix by cceaqm were very close to fox entertaiment.

so I guess it's more of each one preference and what looks better to his eye, i think ...

I'm wondering how cceaqm will function in high compression though ?
the matrix is more dependant on the bitrate moreso than the compression. a movie can have a REALLY high bitrate, so when compressed, the bitrate can still be quite high, but then again, a movie can start with a low bitrate, and when compressed, be even LOWER. but to answer your question the compressions were about 63%, 61%, high 50's and 2 were around 70%.

encoding was done through DVD-RB.

yes, i realize using CCEAQM will use a better matrix, but i'm still wondering why AVAMAT7 looks so good to me, yet the matrix indicates its a lower bitrate matrix, which in theory should eliminate more detail.

[Voodoochild]

the compressions were about 63%, 61%, high 50's and 2 were around 70%.


all of them but the last two, were high compression, maybi that the reason. If the compression is that high, high Matrixes compress less good. In The RME Matrix description, it say High Matrix are good "only if the movie has overhead for compression and if compression is not high".

So in your 50-70% compression maybi lower matrix like AVAMAT7 looks better or softer, but I think to be sure try to do a movie with, lets say 80+ % compression and then compare it..

I'll be glad to know the differences if you do that.

[rahzel]

i guess you're right.

on second look of the movies over 4000 bitrate, using CCE with AQM did look marginally better.

[bennynihon]

You chose a good analogy, but unfortunately it works just the opposite. You could have an audio wave form that was 99.9 % accurate yet would drive everyone listening nuts. All you need for that is a little clipping. And you can have a waveform, that has absolutely no resemblance to the original wave form, yet it would sound allmost identical to the human ear. The later is the case in high bitrate mp3. Due to the loss of some phase relations, the resulting wave looks completely different in time space, but is virtually unchanged in the frequency domain. The former wave form would hardly deviate from the original in the time domain, but show significant deviation in the frequenccy domain. And the human ears sensor array works in the frequency domain.

hey ux-3, I knew when I wrote that, that someone might point that out. As the designer of 3D sound core, I'm aware of this, and really should have said that when comparing all aspects of the audio streams, including the frequency spectrum, and not have kept it as simple as I did.

[Voodoochild]

on second look of the movies over 4000 bitrate, using CCE with AQM did look marginally better.

I think if the movie is not in bad quality, compression is above 70% just using cceaqm is as good as you can get

if movie is high compression, or bad quality or both....
Matrix that compress better and remove more details are better, since some of the noise will not get encoded where matrixes like CCE Default(medium-meium) or avamat 7 is used,

which one to use in those situation? well I think depending on avg bitrate and compression you'll be fine with cce default or avamat7

[ux-3]

hey ux-3, I knew when I wrote that, that someone might point that out. As the designer of 3D sound core, I'm aware of this, and really should have said that when comparing all aspects of the audio streams, including the frequency spectrum, and not have kept it as simple as I did.

I knew, you knew... :p

But then it is a complete comparison, which no longer illustrates the point you tried to make.

This weekend I compared two reductions of lotr3 sse, one with recode, one with cce. In reference to this discussion, I paused the software player frequently. What I found confirms my philosophy of focussing at the running encode, not stills: I was looking at nice stills with both encodes, cce probably better. BUT when I was running them, CCE had some shimmering that was getting on my nerves, and the running movie did not let me catch the details contained. After comparing both encodes at some points, I finally decided to go for the recode2 run, to my own surprise. On the base of stills, I would have decided differently

But here is a dumb question (or a challange). Is there a simple way to create two encodes of the same clip and have them authored as multi angle, so that I can do a side by side comparison at a push of a button? That would be a test I like to do.

Greetings
ux-3

[bennynihon]

What I found confirms my philosophy of focussing at the running encode, not stills

I totally agree that you cannot use stills to accurately compare streams from two different encoders. But when comparing two different quantization matrices with the same encoder, I personally fell that comparing stills is an effective method of comparison....and that is afterall the topic of discussion for this thread.

[Voodoochild]

I'll Be Happy to get some of your conclusions about this post.

CCEAQM=1 base it's matrix on the one you choose as base matrix for the main movie, not just overwrite it.
I extracted all the matrixes for the new encoded DVD with DGIndex, it gave 4 different matrixes all together that AQM used, The most common is the one I chose as the default divided by half!! Meaning if you chose the CCE Default, in most cases it will use matrix that is exactly half of it...
other cases it will use the default, and other it will use matrix that are quarter, of the default...

Examples:

this is CCE Default * 0.75 I used as base Matrix in the "Last Samurai" movie
avg bitrate 3900, compression 71%
used CCEAQM=1,
default matrix for the main movie
CCE Default * 0.75 (the one below)
8 12 14 17 20 20 22 26
12 12 17 18 20 22 26 28
14 17 20 20 22 26 26 29
17 17 20 20 22 26 28 30
17 20 20 22 24 26 30 36
20 20 22 24 26 30 36 44
20 20 22 26 29 35 42 52
20 22 26 29 35 42 52 62
12 13 14 14 15 16 17 17
13 14 14 15 16 17 17 18
14 14 15 16 17 17 18 19
14 15 16 17 17 18 20 20
15 16 17 17 19 20 20 21
16 17 17 18 20 20 21 23
17 17 18 20 20 21 23 23
17 18 19 20 21 23 23 25

that the matrix CCEAQM used most of the time:

8 8 8 9 10 10 11 13
8 8 9 9 10 11 13 14
8 9 10 10 11 13 13 15
9 9 10 10 11 13 14 15
9 10 10 11 12 13 15 18
10 10 11 12 13 15 18 22
10 10 11 13 15 18 21 26
10 11 13 15 18 21 26 31
8 8 8 8 8 8 9 9
8 8 8 8 8 9 9 9
8 8 8 8 9 9 9 10
8 8 8 9 9 9 10 10
8 8 9 9 10 10 10 11
8 9 9 9 10 10 11 12
9 9 9 10 10 11 12 12
9 9 10 10 11 12 12 13

as you can see divided by half.
on some other cases

8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 9
8 8 8 8 8 8 9 11
8 8 8 8 8 9 11 13
8 8 8 8 9 11 13 16
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8

quarter ....

I checked in which cases it uses those Matrixes..
I don't have concrete conclusion.
But in low bitrate lets say below 3000 it uses the quarter Matrix... (lowest number), in other cases where bitrates were between 3000 to 4500 5000 it uses the base Matrix Divided in 2.., and when bitrate were high above 5000 it uses the CCE * .75 which I chose as the base matrix for this project.
Meaning that AQM cut the numbers and make Matrixes with lower values when bitrate drops, to raise quality. When BitRate are high, it uses the default I chose or the one that is divide by 2.

I don't know what you can Make out of this But I will Be glad to here what you think about it.

Sorry for any spelling mistake.
:thanks: