[M4IF Technotes] question about chroma interpolation

[tetsuo kosuge]

Hi Mr.Aggarwal.
I think as follows:
<case1> MB has one MV(1MV) 
	MVchr' = MVlum/2 =(sign)X.Y  (X is integer, Y is decimal)
	uses Table 7-9
	switch(Y){
		0	: Y' = 0;
		0.25	: Y' = 0.5;
		0.5	: Y' = 0.5; 
		1	: Y' = 0.5;
	}
	MVchr = (sign)(X+Y');
<case2> MB has four MV(4MV)
	MVchr' = (MV1lum + MV2lum + MV3lum + MV4lum)/(4*2) = (sign)X.Y (X is integer, Y is decimal)
	uses Table 7-6
	switch(Y){
		0	: Y' = 0;
		0.0625	: Y' = 0;
		0.125	: Y' = 0; 
		0.1875	: Y' = 0.5;
		0.25	: Y' = 0.5
		......(omit).........
		0.8125	: Y' = 0.5
		0.875	: Y' = 1.0
		0.9375	: Y' = 1.0
	}
	MVchr = (sign)(X+Y');
I think that Table 7-7 or 7-8 for the case MB has not 4 luminance block.
Regards,
 Tetsuo
>Can somebody help me understand chroma interpolation, in specific how
>to interpret Tables 7-6 to 7-9. We are trying to implement Simple Profile
>so no quarter-pel. Only half-pel so we need only Table 7-8 and 7-9.
>
>Table 7-8 -- Modification of eighth sample resolution chrominance
>             vector components
>
>eighth pixel position   0  1  2  3  4  5  6  7   //8
>resulting position      0  0  1  1  1  1  1  2   //2
>
>We have 4 luma vectors where each is half-pel. We add them and get a sum
>that is again a max of half-pel resolution. What should we do with this
>sum? Should we first divide by 8 (leading to 16th resolution) and then
>use the table?
>
>What is the meaning of the last column, e.g., //8 and //2 in the
>above table? Can somebody help me with an example for say the sum of
>4 luma vectors = 7, and say 8 and say 14 (where each is half-pel).
>Half-pel Chroma will be 2, 2 and 3. Thus the eventual full-pel chroma
>used for prediction will be 1, 1, and 1, right?
>
>Thanks a lot!
>Gaurav
>
>________________________________________________________________________
>Missed your favourite TV serial last night? Try the new, Yahoo! TV.
>       visit http://in.tv.yahoo.com
>
>_______________________________________________
>Technotes mailing list
>Technotes   lists.m4if.org
>http://lists.m4if.org/mailman/listinfo/technotes
>