On Tue, Nov 08, 2011 at 03:47:31PM +0600, Mashiat Sarker Shakkhar wrote:
> The current code is a generalization of the earlier progressive MV
> prediction code. This was supposed to predict MVs for both interlaced
> and progressive pictures. But the interlaced MV prediction is buggy
> and works mostly by luck.

Nobody likes interlaced modes, eh?
 
> This partially fixes interlaced MV prediction.
> ---
>  libavcodec/vc1dec.c |  189 
> ++++++++++++++++++++++-----------------------------
>  1 files changed, 80 insertions(+), 109 deletions(-)
> 
> diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
> index e389040..3253292 100644
> --- a/libavcodec/vc1dec.c
> +++ b/libavcodec/vc1dec.c
> @@ -1475,12 +1475,10 @@ static inline void vc1_pred_mv(VC1Context *v, int n, 
> int dmv_x, int dmv_y,
>      int px, py;
>      int sum;
>      int mixedmv_pic, num_samefield = 0, num_oppfield = 0;
> -    int opposit, f;
> -    int16_t samefield_pred[2],  oppfield_pred[2];
> -    int16_t samefield_predA[2], oppfield_predA[2];
> -    int16_t samefield_predB[2], oppfield_predB[2];
> -    int16_t samefield_predC[2], oppfield_predC[2];
> -    int16_t *predA, *predC;
> +    int opposit, a_f, b_f, c_f;

nit: oppositE

and while the patch looks reasonably good I wonder if it's not better to split
MV prediction for progressive and interlaced cases
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to