[XviD-devel] Re:Confusion about Interlacing Implemantation

[mayfig]

Hi,
   It is nice to read your e-mail this morning.


   I found some detailed info in the Section of 7.7.2 of MPEG4 Standard last
 
night,and  understand it . Yeah ,I agree with you for 1 and 2 point , if the 

macroblock is a field predicted MB, after decoding, there will be two 

MotionVectors: (MVxf1,MVyf1),(MVxf2,MVyf2) ,which  corresponding to top and 

bottom fields of the MacroBlock. Consequently,in the source code there will be

mb->mvs[0] and mb->mvs[1] which inlude info about the top and bottom fields.

   For the 3 point, I am not sure.Maybe it  depends on the mode of MB, if it is


MODE_INTER4V,then this MB has four MotionVectors; if it is not MODE_INTER4V:

then there are two cases, if the MB is a frame predicted MB, then it has one 

MotionVector,and if the MB is a field predicted MB,then it has two
MotionVectors 

which are for top field and bottom field. 


   But for the three Cases in Section 7.7.2.1, I mean when the
mb->field_pred==1,

which means the current MB is a field predicted MB, I think the implementation

of Xvid has some problem. Moreover , it did not use the info of mb->mvs[1] when


it implement the interpolation of motion vectors.

   One more question, in the decoding process of MB, if  the sequence of MB,

is just 
      for (y = 0; y < mb_height; y++) {
		...
		for (x = 0; x < mb_width; x++) {  
                .....
                }
      }
for the curent MB, how to decide its neighbor MB , i.e. left, top left, and top


right , I do not know how to realize it.
  
   Feel so helpful to discuss it with you , now once again in agreement with
you.

   Thank you for your kind response.


  

_________________________________________________________
Do You Yahoo!?
×¢²áÊÀ½çÒ»Á÷Æ·ÖʵÄÑÅ»¢Ãâ·ÑµçÓÊ
http://cn.rd.yahoo.com/mail_cn/tag/1g/*http://cn.mail.yahoo.com/