Better compression quality YUY2 to MJPEG?

[Sillyname]

When rendering a 3D scene I've been under the impression that if I render to an uncompressed file and then recompress using a lossy codec, I get better quality compression because I now have the frame before and after the current one. Does MJPEG do inter-frame compression or does it literally treat each frame as an individual JPEG picture, ignoring similarities between frames?

------------------
Deep is not the root word of depression.

[Eric VanHeest]

I'm fairly certain that MJPEG, being designed to use as an editing source, only compresses per-frame, so that each frame can be a keyframe. You will always get better compression ratios with a temporal (and intraframe) compression scheme.

[Dr Mordrid]

Yes, MJPeg is an I-frame only compression. So is DV. Still, when editing the addition of effects, overlays or filters introduces DCT (discrete cosine transform) errors because of the re-compression of each frame.

Low data rate MJPeg is notorious for such artifacts. DV is slightly better because of some built-in error correcting features, but it too will show errors after several generations. DV also has a reduced colorspace that can cause problems of it's own when doing effects.

Using raw video (i.e. YUY2, RGB) and having the project settings set up for the same uncompressed format eliminates re-compression errors when adding effects, compositing etc. because there is no DCT compressor involved. In fact sometimes the result looks too good and a blur filter has to be added to make it look more "real".

Once the project is finished then you can render the whole shebang to a temporal format or create it as a raw video file for use with an external compressor. Either will usually give you very high quality.

Dr. Mordrid