x264_compare

[pokkys]

There is a little test.

Sorry,I can't upload all the sample.

For Some pic shows what session it is.

Fixed quantizer : 26

b for b-frame,r for ref

Fast-Motion session:
b0_r1 6.77MB 100%
b0_r5 6.46MB 95.4%
b0_r15 5.52MB 81.5% <-1
b1_r1 6.64MB 98.1% <-2
b1_r5 5.68MB 98.1%
b1_r15 5.39MB 79.6%

Slow-Motion session:
b0_r1 3.65MB 100%
b0_r5 3.58MB 99.1%
b0_r15 3.46MB 95.8% <-3
b1_r1 3.44MB 99.1% <-4
b1_r5 3.34MB 95.7%
b1_r15 3.25MB 94.0% <-5

1. 15 ref is still good in fast-motion.
2. b-f is sometimes better than adding more ref.(maybe in another test)
3. Adding ref works little in Slow-Motion.
4. b-f is the same as above.
5. Slow-Motion is already good for compressing without b or r.
*6. This case may be "uni-case" or "my-case".

[akupenguin]

Note that you can't simply use filesize at fixed quantiser to compare B-frames: the quality will be different.