HCenc 023 released

[manolito]

I did a couple of test encodes using either *AQ 3 or *LUMGAIN 3. The source was quite dark, average bitrate a little over 5000 kbps. The matrix I used comes from manono:
8, 8, 8, 9,11,13,14,17,
8, 8, 9,11,13,13,14,17,
8, 8,11,12,13,14,17,94,
9,11,13,13,14,17,17,94,
11,11,13,13,14,17,94,94,
13,13,14,16,17,20,94,94,
13,13,14,17,94,94,94,94,
13,14,17,94,94,94,94,94

12,12,13,14,15,16,22,26,
12,13,14,15,16,22,26,32,
13,14,15,16,22,26,32,41,
14,15,16,22,26,32,41,53,
15,16,22,26,32,41,53,94,
16,22,26,32,41,53,70,94,
22,26,32,41,53,70,94,94,
26,32,41,53,94,94,94,94


To make it short: I could not detect any differences between the two encodes. Maybe the bitrate was too high. AQ 3 did compress slightly better than LUMGAIN 3, but this probably depends on the source.

So more testing needed, I guess...

Cheers
manolito

[Boulder]

Try sources with flat areas versus sharper edges..the bits should be shifted from the edges towards the more flat areas of the frames. Also scenes with grass or similar texture are quite good.

[blutach]

As usual, I am late with my thanks Hank.

Regards

[asarian]

Outstanding piece of work, that HC encoder of yours!

[phXql]

Great piece of software. Thank you!

[six13]

Can someone please explain the 4 settings in AQ vs the 4 settings in LUMGAIN? I am trying to understand what to look for. I have previously used LUMGAIN 4 and liked the video. AQ has me puzzled, I saw the comments of Boulder but am still puzzled. I mainly compress/encode TV captures movies at a bitrate of 4K/S with small compression % like 82% with DVD-RB. I do notice that the detail on a face isn't real good it looks kinda flat. I am learning and would welcome opinions. Thanks in advance.

[Dark Shikari]

Originally Posted by six13

I do notice that the detail on a face isn't real good it looks kinda flat.

Unless I'm mistaken, HC uses luminance-masking adaptive quantization, which won't help in such a case (one needs complexity masking like x264 has).

[Boulder]

AQ adapts the quantization, LUMGAIN affects the quantization matrix. hank posted the code related to LUMGAIN in the HC Encoder thread which shows how it works. Basically LUMGAIN lowers the coefficients of the base quantization matrix depending on the luminance levels. The strength affects the factor used in determining the new coefficients.

If I'm not entirely mistaken, AQ is meant for shifting bits inside one frame - from sharp edges to the more flat areas. LUMGAIN is supposed to shift bits to low-luminance frames from the brighter ones to prevent blocking in dark frames.

[six13]

So Boulder, then is it OK to use them at the same time or is it best to pick 1 or the other? In the past you have sugested I use LUMGAIN 4 and i like the output but was wondering if AQ would produce better results then LUMGAIN. What I recently encoded didn't have many dark sections. What do you use to make the decision as to use LUMGAIN or AQ?

[Boulder]

I use 3 for both AQ and LUMGAIN. They are two different things so they can be used together.

[hank315]

Originally Posted by Boulder

AQ is meant for shifting bits inside one frame - from sharp edges to the more flat areas. LUMGAIN is supposed to shift bits to low-luminance frames from the brighter ones to prevent blocking in dark frames.

Yes, that's what it does. The final quant for each MacroBlock: Q(i,j) = QMB * Qmatrix(i,j).
QMB is adjusted by the variance of the MacroBlock --> AQ.
Qmatrix(i,j) is adjusted by the average luminance of the frames in a GOP --> LUMGAIN.

Originally Posted by Dark Shikari

Unless I'm mistaken, HC uses luminance-masking adaptive quantization, which won't help in such a case (one needs complexity masking like x264 has).

HC doesn't use luminance-masking adaptive quantization, the quantization is adapted by the variance of the MB so it's some kind of a complexity mask.
This picture shows the Quants of the same frame for different AQ values.

[Dark Shikari]

Originally Posted by hank315

HC doesn't use luminance-masking adaptive quantization, the quantization is adapted by the variance of the MB so it's some kind of a complexity mask.
This picture shows the Quants of the same frame for different AQ values.

Oh, so its like VAQ, complexity masking. Good to know that you've implemented a useful AQ method

[six13]

Thanks Boulder, which matrix do you use with those settings in HCenc?

[Boulder]

The matrix depends entirely on the source. Use the search, there are quite a few threads about this issue.

[zeropc]

hank, first let me say this is a very nice tool. i just started using it and the results are so far pretty good. but there are some open question for me in fully understanding the usage and some minor issues. i hope you or somebody else can help with this.

1. i currently set aq to 0 and the luminance gain to 0 , is this good or should i raise it? i encode from br/hd-dvd sources at 18mbps
2. i experienced that hcenc seams to lower the colors by a bit. i re-encoded the simpsons movie and saw how the skin colors faded a bit, which kinda bugs me
3. is there a way to scale hd sources to dvd res (pal and ntsc)?
4. is it possible to import .dga files from dgavcindex besides using avisynth?

[gizzin]

1. I say leave it on, especially at the bitrate you are using.
2. I never heard of that before, but maybe, how/what program are you using, what kind of file (DVD,XVID), essentially what I'm saying is you need to provide more information.
3. Yes
4. Not sure. I wouldn't think so.

Hank, I get this mismatch error with v23. "Error, large source mismatch found in pass 2 starting at frame:0, count :1 frames." I'm using FAVC 1.06. Just thought I'd point that out. It seems to be only this program that experiences this problem, as I use DVD Rebuilder and it never happened.

[Graigddu]

gizzin i remember that i came across this error before or something similar and after a bit of searching found that ticking reload avisynth during second pass under the settings 3 tab in the HCGUI fixed the problem, remember to save HC Ini in main tab or any settings changes won't take effect

[Revgen]

Quality is good as long as I keep AQ off. Not much different than HC022.

[BDrift]

Zones don't seem to work for me

*ZONE 3
1 0.2
1200 1.0
163500 0.2

Thats what I use in hc.ini, but when I compare it with a encode without zones, there is close to zero difference visible, not even in a frame by frame comparison. Am I missing something? Shouldn't frames 1-1200 and frames 163500-end have a ~80% lower average bitrate?

For testing I've set end frame to 2000 ... could that be causing trouble?

[lithoc]

I've found a bug that cause Ifoedit to crash when muxing.

I encoded some 352x240 video. I found out that Ifoedit will crashed at the beginning if the video bitrate is below ~256kbps.

I notice video having a black fade in. When the black is usually low bitrate that's where the Ifoedit crashed.

If I encode the video starting frames of 100 where there's no black screen. Ifoedit mux just fine when the bitrate is > 256kbps.

Can HCenc have minimum bitrate settings?