[x264-devel] Question for intra prediction

[Taeyoung Na]

Dear all..

Sorry for the first mail. It was my first question so I couldn't think for using plain text mode.

I’m analyzing intra prediction part in analyse.c

There, I feel so strange for the use of i_pred_mode variable.

For 4x4 intra prediction in x264_mb_analyse_intra() function , we can see below code.

if( b_merged_satd && i_max >= 6 )
{
int satd[3];
         h->pixf.intra_satd_x3_4x4( p_src_by, p_dst_by, satd );

	/* this is the part where I want to know ...........
         if( i_pred_mode < 3 )
                satd[i_pred_mode] -= 3 * a->i_lambda;
         for( i=2; i>=0; i-- )
         COPY2_IF_LT( i_best, satd[i] + 4 * a->i_lambda, a->i_predict4x4[idx], i );
         i = 3;
	..........................................*/ 

}
else
         i = 0;
.......................
.......................

i_pred_mode is defined by the below line.

int i_pred_mode = x264_mb_predict_intra4x4_mode( h, idx );

this value could have mode 0(vertical) to 8(horizontal up)

Here I have a question, “Why the encoder try to decide a best intra mode same as the i_pred_mode?”        - See /*.....................*/ above code

There is no time saving effect, no PSNR preservation as far as I know.

if the QP is 26, a->i_lambda is 25 so satd[i_pred_mode] = -75 initially (other modes have 0 initially).

Although it’s a small value for the mode decision but I just wanna know the reason.

Looking forward to hearing from ones who can show me the reason soon….

Sincerely.




----------------------------------------------------------------------------
On Mon, 25 Jun 2007, 3*EB?5 wrote:

> Here I have a question, !0Why the encoder try to decide a best intra 
> mode among the mode 0,1,2(Vertical, Horizontal, DC) ?!1 ),- See the 
> red part of the code

HTML email is not welcome here. Please write any formatting relevant to your content in plain text, and please omit any irrelevant formatting.
In case that's not clear: no, I don't see any red part, and I don't know what you're asking.

--Loren Merritt



_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel