[mythtv-users] 1080i to 1080p Deinterlacing on Backend was Raspberry Pi now ships with 512MB RAM

[Tim Phipps]

Quoting Joseph Fry <joe at thefrys.com>:

> If you were to pre-deinterlace a file, you would be transcoding that file
> in the process.  Essentially, you would render a frame, deinterlace it, and
> recompress it.  So lets say you have the best deinterlacer you could
> possibly imagine and get the best possible resulting progressive frame...
> now you need to compress it to the new video file... here is where your
> efforts are wasted... whenever you recompress a video your going to
> introduce artifacts.  In the end the resulting video may actually be of
> lower quality than your current interlaced video and deinterlacer.

It may not be that bad as long as you can link the decompression to  
the recompression. What I mean is one big program that does the lot  
rather than those lovely unix pipes. You can recompress video  
losslessly as long as you use the same DCT coefficients. In this case  
the deinterlace step is going to muck around with the uncompressed  
pixels so there will be some data loss but the most mucking around  
happens with moving objects and you are less likely to notice  
artefacts the faster a thing is moving. What you will end up with  
though is a file with twice the number of uncompressed pixels so it  
will be quite a but bigger on disk.

Obviously - writing the above program is left as an exercise for the reader.

Cheers,
Tim.