Fast-pskip | No-fast-pskip | ||
Burn | SSIM | 0.816305 | 0.809181 |
rate | 603.07 | 605.67 | |
FiM intro | SSIM | 0.970886 | 0.970532 |
rate | 622.86 | 625.07 | |
Rush hour | SSIM | 0.897251 | 0.901464 |
rate | 599.98 | 601.09 | |
Students | SSIM | 0.986635 | 0.986645 |
rate | 599.26 | 599.33 | |
Burn | SSIM | 0.722376 | 0.712578 |
rate | 294.58 | 298.1 | |
FiM Intro | SSIM | 0.937741 | 0.934879 |
rate | 320.88 | 324.85 | |
Rush hour | SSIM | 0.854779 | 0.858797 |
rate | 305.84 | 308.25 | |
Students | SSIM | 0.979623 | 0.979623 |
rate | 300.39 | 300.58 |
Four test videos were used: Three from the Xiph.org test-video library, and the opening title of Friendship is Magic as an animation comparison. Each of these was encoded four times: Once at 'standard' x264 defaults, one at the same with the addition of no-fast-pskip, once at 'extreme' settings as might be used when trying to encode for very low bitrates and once at the same settings with the addition of no-fast-pskip. Overall, sixteen encodes were performed: Four videos, each under four x264 settings. All encodes were performed under Scientific Linux using mencoder SVN-r34833-snapshot-4.4.5 and x264 core 122.
All but one of the videos both an increase in bitrate and a lowering of SSIM when no-fast-pskip was added. The only video in which SSIM increased with the addition of no-fast-pskip was the Rush Hour video, when the slight increase can be dismissed as a consequence of the increased bitrate. After these tests, I conclude that no-fast-pskip is of little practical use.