Re: Hdot264-devel digest, Vol 1 #29 - 2 msgs

"yulejun" <[email protected]> Thu, 17 Apr 2003 21:57:10 +0000
Newsgroups gmane.comp.video.h264.devel
Message-ID <[email protected]>
In jvt-b038.doc, qbits = 17 + QP/6.
And In the JM5.0c, ABT_N=17 (see abt_const.c). I noted that 12 is subtracted 
from QP. 

Set qp = QP -17 

so qbit = 17 + qp/6 = 15 + QP/6 

But I do not know how the a^2 ab/2  b^2/4 are counted in JM6 

 


[email protected] : 

> Send Hdot264-devel mailing list submissions to
> 	[email protected] 
> 
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://lists.sourceforge.net/lists/listinfo/hdot264-devel
> or, via email, send a message with subject or body 'help' to
> 	[email protected] 
> 
> You can reach the person managing the list at
> 	[email protected] 
> 
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Hdot264-devel digest..." 
> 
> 
> Today's Topics: 
> 
>    1. confusions about qbits and quant coef matrix (Eric Miao)
>    2. Re: confusions about qbits and quant coef matrix (Aitor Garay) 
> 
> --__--__-- 
> 
> Message: 1
> Date: Tue, 15 Apr 2003 06:18:23 -0700 (PDT)
> From: Eric Miao <[email protected]>
> To: [email protected]
> Subject: [Hdot264-devel] confusions about qbits and quant coef matrix 
> 
> --0-305635399-1050412703=:68618
> Content-Type: text/plain; charset=us-ascii
> Content-Id: 
> Content-Disposition: inline 
> 
> Hi, 
> 
> In jvt-b038.doc, qbits = 17 + QP/6.
> But I also noted that in some articles and jm6 ref
> software, qbits = 15 + QP/6.
> What is right? 
> 
> And I also calculated quant coef matrix according to
> equation: MF/qbits = PF/qstep.
> My result is 
>  a^2   (b^2)/4 (ab)/2
> 13107     5243     8290
> 11916     4766     7536
> 10082     4033     6377
>  9362     3745     5921
>  8192     3277     5181
>  7282     2913     4605
> But it is different from the one used in jm61
> software,
> mainly column 2 and column 3.
> The following is used in jm61:
>  a^2   (b^2)/4 (ab)/2
> 13107 5243 8066
> 11916 4660 7490
> 10082 4194 6554
>  9362 3647 5825
>  8192 3355 5243
>  7282 2893 4559 
> 
> Attached is a c file I used to calculate the matrix.
> Can anyone point out any wrong in my implem? 
> 
>  
> 
> __________________________________________________
> Do you Yahoo!?
> The New Yahoo! Search - Faster. Easier. Bingo
> http://search.yahoo.com
> --0-305635399-1050412703=:68618
> Content-Type: text/plain; name="calc_mf.c"
> Content-Description: calc_mf.c
> Content-Disposition: inline; filename="calc_mf.c" 
> 
> #include <stdio.h>
> #include <math.h> 
> 
> static double quantStep[6] =
> {
>     0.625,
>     0.6875,
>     0.8125,
>     0.875,
>     1.0,
>     1.125
> }; 
> 
> static double postScalingFactor[3]; 
> 
> int
> main()
> {
>     double a, b;
>     int i; 
> 
>     a = 0.5;
>     b = sqrt(0.4); 
> 
>     postScalingFactor[0] = a * a;
>     postScalingFactor[1] = (b * b) / 4;
>     postScalingFactor[2] = (a * b) /2; 
> 
>     for(i = 0; i < 6; i++)
>     {
>         printf("%8d %8d %8d\n",
>                (int)((postScalingFactor[0] * 32768) / quantStep[i] + 0.5),
>                (int)((postScalingFactor[1] * 32768) / quantStep[i] + 0.5),
>                (int)((postScalingFactor[2] * 32768) / quantStep[i] + 0.5));
>     } 
> 
>     return 0;
> } 
> 
> --0-305635399-1050412703=:68618-- 
> 
> 
> --__--__-- 
> 
> Message: 2
> From: "Aitor Garay" <[email protected]>
> To: "Eric Miao" <[email protected]>,
> 	<[email protected]>
> Subject: Re: [Hdot264-devel] confusions about qbits and quant coef matrix
> Date: Tue, 15 Apr 2003 16:16:58 +0200 
> 
>     In the reference software the positions with factors (b^2)/4 and ab/2 have
> been modified, but i ignore the rationale they followed to do so.  It's OK to
> modify the forward quantizer since it is not standarised, only the rescaling
> process is binded to the standard. 
> 
>     /AITOR 
> 
> ----- Original Message ----- 
> From: "Eric Miao" <[email protected]>
> To: <[email protected]>
> Sent: Tuesday, April 15, 2003 3:18 PM
> Subject: [Hdot264-devel] confusions about qbits and quant coef matrix 
> 
> 
>> Hi, 
>> 
>> In jvt-b038.doc, qbits = 17 + QP/6.
>> But I also noted that in some articles and jm6 ref
>> software, qbits = 15 + QP/6.
>> What is right? 
>> 
>> And I also calculated quant coef matrix according to
>> equation: MF/qbits = PF/qstep.
>> My result is 
>>  a^2   (b^2)/4 (ab)/2
>> 13107     5243     8290
>> 11916     4766     7536
>> 10082     4033     6377
>>  9362     3745     5921
>>  8192     3277     5181
>>  7282     2913     4605
>> But it is different from the one used in jm61
>> software,
>> mainly column 2 and column 3.
>> The following is used in jm61:
>>  a^2   (b^2)/4 (ab)/2
>> 13107 5243 8066
>> 11916 4660 7490
>> 10082 4194 6554
>>  9362 3647 5825
>>  8192 3355 5243
>>  7282 2893 4559 
>> 
>> Attached is a c file I used to calculate the matrix.
>> Can anyone point out any wrong in my implem? 
>> 
>>  
>> 
>> __________________________________________________
>> Do you Yahoo!?
>> The New Yahoo! Search - Faster. Easier. Bingo
>> http://search.yahoo.com
>  
> 
> -------------------------------------------------------------------------------- 
> 
> 
>> #include <stdio.h>
>> #include <math.h> 
>> 
>> static double quantStep[6] =
>> {
>>     0.625,
>>     0.6875,
>>     0.8125,
>>     0.875,
>>     1.0,
>>     1.125
>> }; 
>> 
>> static double postScalingFactor[3]; 
>> 
>> int
>> main()
>> {
>>     double a, b;
>>     int i; 
>> 
>>     a = 0.5;
>>     b = sqrt(0.4); 
>> 
>>     postScalingFactor[0] = a * a;
>>     postScalingFactor[1] = (b * b) / 4;
>>     postScalingFactor[2] = (a * b) /2; 
>> 
>>     for(i = 0; i < 6; i++)
>>     {
>>         printf("%8d %8d %8d\n",
>>                (int)((postScalingFactor[0] * 32768) / quantStep[i] + 0.5),
>>                (int)((postScalingFactor[1] * 32768) / quantStep[i] + 0.5),
>>                (int)((postScalingFactor[2] * 32768) / quantStep[i] + 0.5));
>>     } 
>> 
>>     return 0;
>> } 
>> 
>  
> 
>  
> 
> --__--__-- 
> 
> _______________________________________________
> Hdot264-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/hdot264-devel 
> 
> 
> End of Hdot264-devel Digest
 

 ----------------------------------------------------------------
FESCOmail.net


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf