What is a High/medium/low bitrate?

[six13]

What figures to consider for high/medium/low bitrates. I just encoded a TV capture standard deff and this is what DVD-RB gave me for phase 1:

18:47:03] Phase I, PREPARATION started.
- VTS_01: 1,650,184 sectors.
-- Scanning and writing .D2V file
-- Processed 168,899 frames.
-- Building .AVS and .ECL files
- VTS_02: 820,311 sectors.
-- Scanning and writing .D2V file
-- Processed 84,019 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 90.0%
- Overall Bitrate : 3,291Kbs
- Space for Video : 4,237,942KB
- HIGH/LOW/AVERAGE Cell Bitrates: 3,332/3,266/3,291 Kbs
[18:49:16] Phase I, PREPARATION completed in 2 minutes.

I am playing with diff. matrix that are in HCenc and some I have found on this board. I just used Auto-Q1.mtx

With my above cell bitrates which Matrix should I use, I have just used the standard MPG from HCenc for everything in the past. I know it is a tough question to answer since what one person like otheres may not. Any tips on Matrixs to use would be appreciated for my SD TV captures which are done with SVIDEO. I really ask since I see in the matrix names things like medium bit rate, or low bitrte or high bitrate matrix. What constitutes low, medium, high?

Thanks in advance.

[kumi]

I'm not an expert on this, so I'll let someone else answer the meat of your question.

However, you should also look at the percentage compression value: 3,291Kbs may be a 'low' bitrate for a source compressed at 65%, but at 90% you're not losing much to begin with.

[Boulder]

Hi/med/low bitrate depends entirely on your source - how many bits it needs to yield a result that pleases your eyes.

[six13]

OK, but if I am considering a Matrix to use. So when they refer to a high or low bitrate matrix they refering to the bitrate of the source? So for my example in post #1 with a compression rate of 90% for DVD5 and HIGH/LOW/AVERAGE Cell Bitrates: 3,332/3,266/3,291 Kbs naturally this is a NTSC TV capture so the bitrate will be lower than a DVD rip since I am so close to the original at 90% would I use a high bitrate matrix or since the average bitrate is 2,291 Kbs would I select a medium or low bitrate matrix since it is in the 3,000 range?

[Boulder]

It's still pretty much the same principle. Whenever I need to think about having a high or low bitrate, I encode the clip with the MPEG standard matrix with CCE, doing an OPV at Q30. Then I compare the resulting avg bitrate to the bitrate that I can allocate and choose the matrix according to that. In my eyes, Q30 with the MPEG standard matrix produces satisfactory results. That is why I use it as a baseline for the analysis.

I can't state any absolute truths though. A small rule of thumb in my encodes is that if the test encode yields a lower avg bitrate than what I could allocate, I use a higher bitrate matrix than the MPEG standard and vice versa. If the avg bitrates are close to each other, I use the MPEG standard.

The issue is far from simple, IMO you need to develop your own methods for deciding what matrix to choose..or just choose the MPEG standard and enable adaptive quant matrices (in rebuilder.ini, CCEAQM=1 for CCE and the HC_LUMGAIN parameter for HC, I use HC_LUMGAIN=4).

[six13]

Thanks Boulder,

I have read that you were using LUMGAIN 4. Corect me if I am wrong but itsn't that feature really for dark parts of the video. My source is a sporting event shot outside, do you still think it would be worth chosing LUMGAIN 4 in the HCenc GUI for this source? would it help with compression? I think jdobbs said he liked 2 for LUMGAIN as, I remember reading about the feature with udpdate 21 for HCenc when you both were speaking about it.

thanks

[Boulder]

You can test the feature by encoding with LUMGAIN 2 or 4 and then opening the video in DGIndex and choosing to log the quant matrices. You'll see that the matrices are adapted constantly (at least with 4) so it's not applied only to dark scenes.

[six13]

Thanks, I found a Sony medium bitrate matrix posted on this board and used it with LUMGAIN 4 and my results are incredible! for my TV capture (4000 KBS). Not too smooth, just incredible, the poster had a Sony high bitrate version also and I put it in the matrix folder also. I can't find it now on the board. I started searching for the high quality Fox matrix and came across the 2 Sony ones.

[six13]

I see when I load my custom matrix in the HCnc GUI and select LUMGAIN = 4 the message in the HCenc GUI says luminanace gain active, adaptive matrix disabled. What does this phrase adaptive matrix disabled mean? The term adaptive has me confused since it still must be using the chosen custom matrix for encoding since it needs a matrix.

[45tripp]

You'll see that the matrices are adapted constantly (at least with 4) so it's not applied only to dark scenes.
Adapted matrices,
adaptive matrices.
Lumigain is adapting the base matrix based on source luminance.

In the matrix tab, top right is a check box for enabling an adaptive matrix using a matrix.dll you can create.
Adaptive matrix code can change the matrix for every group of pictures.
It's this that's disabled, as with lumigain you are already adapting the matrix. Can't have them both.

[kumi]

Indeed, enable HC's log to see that adaptive is on when lumgain > 0.

[six13]

OK, thanks. I tried LUMGAIN = 2 but liked the output from 4 better. Not too smooth, not to sharp. With my previous method the picture was pretty harsh at 65%-90% compression # i even tried fluxsmooth or removegrain but didn't really help. I prefered HCenc at 53% better for my TV capture since the output was smoother looking and not as harsh or sharp.

This Sony medium bitrate matrix with LUMGAIN 4 at 70% compression looks very good for a TV capture. The diference is very obvious.

I am learning new stuff every day. :thanks: