.

.
.
.
.
.
.
.
. scroll down
.
.

Monday, July 26, 2010

2 pass encoding vs. constant quality encoding

2010.7.28 15.00
and why frame grabs are not always reliable when you want to compare quality.

I tried to grab exactly the same frame every time what proofed to be futile, so this is as close as its get.
Also I made lossless crops from the 1280x720 screen shot  in 640 x 640 pixel that you do not need to click on the picture to see the full size. This size is the maximum Blogger can display in the posting. They are re-compressed by Blogger, but I compared the posting images with my originals and they are comparable.
All grabs are done in GOM player.
Just hit F11 now to go full screen in your browser and it is very easy to just scroll up and down to fast compare the results. When you are finished hit F11 again and your browser is back to normal.

RF:25 190 MB  1100 Kbps



2 pass 194 MB  1130 Kbps

2 pass 194 MB made another grab after the one above looked too bad

RF:28 130 MB  715 Kbps

2 pass 133 MB  732 Kbps

RF:30.5 104 MB  574 Kbps

2 pass 102 MB  565 Kbps

2 pass 100 MB  after the grab above looked so bad I loaded the clip again and tried another one and that one looked much better.

RF:33 75 MB  420 Kbps

2 pass 75 MB  431 Kbps

***

After seeing the differences in the different encodings, try to figure out with what method and what bitrate was used in the next 3 samples.

I

II

III

So, did you make up your mind?
.
.
.
.
.
.
.
then
.
.
.
.
.
.
.
scroll
.
.
.
.
.
.
.
.
down
.
.
.
.
.
.
.
.
.
you guessed wrong!

I, II and II are from the same original AVCHD encoding what had a size over 550 MB and a bitrate of 3300 Kbps.

Bummer!

So what can we learn about that?
Screen grabs can only tell you so much. If you grab a few frames (a few milliseconds) later, the imperfections might already be fixed and your comparison will be flawed.
So it is best to view the video on the big screen, find the areas what has problems and compare these time frames during normal play back if they bother at all.

This reminds me a lot of pixel peeping at digital cameras. The final result what you can make from the raw material is what counts.


What I learned from this little exercise is: Stick with RF:25 and if I need to really squeeze a video then go to 2 pass encoding.

---

Here is a comparison between a 22 min South Park episode
550 MB standard size encoding and a RF:25 encoding what resulted in a 98 MB file.
Both use the same audio compression.

550 MB

98 MB


For my eyes encoding files that large makes no sense, except to waste a lot of space on the MMPC drive.

1 comment:

  1. :) Google for I, P and B frames, then you'll understand why. Always compare P-frames, coz I-frames are not relevant.

    ReplyDelete