[XviD-devel] Decoding B and P frames

[Stephan Assmus]

Hi,

reno reballos wrote:
> another thing i want to ask if you know about this with regards to 
> decoding a compressed frame. why decoder could not decode them one time? 
> i mean decoder will have to execute again to decode the remaining 
> compressed frame... what is the reason behind...?

Ah, I think I may know what the problem is: This has been discussed here 
very recently: The decoder needs to be fed with complete frames, which is 
normally taken care of when the stream is embedded into an AVI container 
for example. So in your case, I assume you are dealing with the raw xvid 
stream and read fixed size buffers. I believe the frames inside the stream 
are arbitrarily broken up by the fixed buffer sizes and that is your 
problem. You need to embed "markers" into the stream. I mean, the stream 
already contains MPEG4 markers for frames, but it will be easier if you 
write some markers yourself. For example, you could write the size of the 
next chunk of encoded data. Right now, I assume you do the following:

1) encode a frame
2) write encoded data to disk
3) goto 1) while frames are left

Replace this with this:

1) encode a frame
2) write a 32 bit value with the size of the encoded data to disk
3) write encoded data to disk
4) goto 1) while frames are left

Your decoding loop probably looks like this:

1) read fixed size buffer of data from disk
2) feed data into decoder
3) goto 1) while not yet encoded a frame
4) display frame
5) goto 1) while data left

Replace this with this:

1) read 32 bit value from disk
2) create buffer with that size
3) read entire buffer from disk
4) feed data into decoder
5) goto 1) while not yet encoded a frame
6) display frame
7) goto 1) while data left

This way, you simulate the buffers with chunk size which you would get from 
containers like AVI. Note that 5) is still necessary!

Hoping my guess is right at what your problem is and also hoping this makes 
it clear what I envision for a solution.

Best regards,
-Stephan