DeathTheSheep
18th February 2005, 03:11
Not sure if this is the best place to put this, but here goes:
B-frames are special frames made by estimating what a frame should be based on those around it. Thus they conserve bitrate. However, if there was a way to do a similar motion prediction without actually adding any frames at all, they wouldn't be as necessary, because the decoder would insert them in dynamically.
WMV9 used to utilize a decoder-side motion smoother for lower-quality wmv9 clips less than 15fps. By doing this, it emulated the frame rate of the original source despite the fact that there was actually no data in the video stream to create these extra frames.
Due to the variety of decoding software available, such an endevor wouldn't be as feasable as one might hope. However, there is a solution--integration into ffdshow or availability in the form of a filter (motionSmootherFilter), which would boost the framerate by inserting any number of predicted frames in between each P-frame in realtime, thereby ensuring excellent motion smoothing at lower bitrates.
If incorporated as a filter, its functionality could be extended further with the ability to smooth existing motion (rather than merely estimating and inserting frames), making it extremely useful in anime, where all character motion is fairly choppy due to its "animated" nature.
That's the theory, so where's the filter? Is it something I've managed to overlook? Or is it far more complex than it is made out to be, despite the fact that MS somehow managed to do it?
B-frames are special frames made by estimating what a frame should be based on those around it. Thus they conserve bitrate. However, if there was a way to do a similar motion prediction without actually adding any frames at all, they wouldn't be as necessary, because the decoder would insert them in dynamically.
WMV9 used to utilize a decoder-side motion smoother for lower-quality wmv9 clips less than 15fps. By doing this, it emulated the frame rate of the original source despite the fact that there was actually no data in the video stream to create these extra frames.
Due to the variety of decoding software available, such an endevor wouldn't be as feasable as one might hope. However, there is a solution--integration into ffdshow or availability in the form of a filter (motionSmootherFilter), which would boost the framerate by inserting any number of predicted frames in between each P-frame in realtime, thereby ensuring excellent motion smoothing at lower bitrates.
If incorporated as a filter, its functionality could be extended further with the ability to smooth existing motion (rather than merely estimating and inserting frames), making it extremely useful in anime, where all character motion is fairly choppy due to its "animated" nature.
That's the theory, so where's the filter? Is it something I've managed to overlook? Or is it far more complex than it is made out to be, despite the fact that MS somehow managed to do it?