[FFmpeg-user] video to image sequence conversion problem
samosa_chaat at yahoo.com
Mon Jul 9 20:31:43 CEST 2012
>> My video (29.97 fps) is that of a precise oscillatory motion
>> of time period 40 seconds. If the downsampling to 5 fps is
>> done correctly then the peaks (in motion) should be 200
>> (=5*40) downsampled frames apart.
>Aren't you assuming 30fps here?
>> But I find that it is off by 1 or 2 frames in the downsampled
>> sequence (a difference of about 6-12 original frames).
> Isn't this exactly the difference between 30000/1001 and 30fps?
I'm not assuming 30fps in those calculations. Even if the recorded video is x fps (say x is larger than 5) and I downsample to 5 fps, then a harmonic motion of 40 seconds should be covered in 200 frames.
You are right that the difference seems to be the same as 30000/1001 and 30fps. But why is this a factor in image sequence conversion? Cinelerra doesn't have this problem.
Maybe I need to backup and ask how the downsampling is done. Is it done by skipping every n frames (where n is an integer) or is it done by constructing (interpolating/extracting) frames at exact time points? If it's the latter then the image-time relation should be in perfect sync irrespective of the input fps (30 or 29.997), correct?
More information about the ffmpeg-user