render.c

来自「VLC媒体播放程序」· C语言 代码 · 共 1,920 行 · 第 1/5 页

/*****************************************************************************
 * render.c : Philips OGT and CVD (VCD Subtitle) blending routines.
 *            stuff from here might be pulled out, abstracted or used
 *            by DVD subtitles.
 *****************************************************************************
 * Copyright (C) 2003, 2004 VideoLAN
 * $Id: render.c,v 1.29 2004/02/02 12:53:20 fenrir Exp $
 *
 * Author: Rocky Bernstein <[e-mail]>
 *   based on code from: 
 *          Sam Hocevar <[e-mail]>
 *          Rudolf Cornelissen <[e-mail]>
 *          Roine Gustafsson <[e-mail]>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
 *****************************************************************************/

/*#define TESTING_TRANSPARENCY 1*/

/*****************************************************************************
 * Preamble
 *****************************************************************************/
#include <vlc/vlc.h>
#include <vlc/vout.h>
#include <vlc/decoder.h>

#include "pixmap.h"
#include "subtitle.h"
#include "render.h"

/* We use 4 bits for an transparency value: 0..15, 15 is completely
   transparent and 0 completely opaque. Note that although SVCD allow
   8-bits, for these routines pixels should previously have been be
   scaled down to 4 bits (the range used by DVDs).
*/
#define TRANS_BITS (4)
#define MAX_TRANS  ((1<<TRANS_BITS) - 1) 
#define TRANS_SCALEDOWN (8-TRANS_BITS)

/* We use a fixed-point arithmetic in scaling ratios so that we
   can use integer arithmetic and still get fairly precise
   results. ASCALE is a left shift amount. 
*/
#define ASCALE 6  /* 2^6 = 32 */

/* Horrible hack to get dbg_print to do the right thing */
#define p_dec p_vout

/**
   Take two 8-bit RGB values and a transparency and do a weighted
   average of them. The "weight" comes from i_trans which is in the
   range 0..MAX_TRANS. To have greater precision using integer
   arithmetic, the RGB values are scaled. The uint16_t cast below is
   to make sure we don't overflow the product in the multiplication

   (MAX_TRANS - i_trans) is the additive "inverse" of i_trans, i.e.
   i_trans + (MAX_TRANS - i_trans) = MAX_TRANS. So the resulting sum
   of rgb1*i_trans + rgb2*(MAX_TRANS-i_trans) will be scaled by
   MAX_TRANS. To reduce the value back down to 8 bits, we shift by
   TRANS_BITS, noting that 1 << TRANS_BITS is roughly
   MAX_TRANS. (Actually it is MAX_TRANS - 1).
*/
#define avg_8bit_rgb(rgb_vout, rgb_sub, i_trans)                   \
{                                                                  \
   int i;                                                          \
   for (i=0; i < RGB_SIZE; i++) {                                  \
   rgb_vout[i] = ( (uint16_t) rgb_vout[i]*(MAX_TRANS-i_trans) +    \
                   (uint16_t) rgb_sub [i]*i_trans ) >> TRANS_BITS; \
   }                                                               \
}


/*****************************************************************************
 * Local prototypes
 *****************************************************************************/
static void BlendI420( vout_thread_t *, picture_t *, const subpicture_t *,
                        vlc_bool_t );
static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop );
static void BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
                       const subpicture_t *p_spu, vlc_bool_t b_crop,
                       vlc_bool_t b_15bpp );
static void BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop );
static void BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop );
static void BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop );

/*****************************************************************************
 * BlendSPU: blend a subtitle into a picture
 *****************************************************************************
 
  This blends subtitles (a subpicture) into the underlying
  picture. Subtitle data has been preprocessed as YUV + transparancy
  or 4 bytes per pixel with interleaving of rows in the subtitle
  removed. 

 *****************************************************************************/
void VCDSubBlend( vout_thread_t *p_vout, picture_t *p_pic,
		   const subpicture_t *p_spu )
{
    struct subpicture_sys_t *p_sys = p_spu->p_sys;
  
    dbg_print( (DECODE_DBG_CALL|DECODE_DBG_RENDER), 
	       "chroma %x", p_vout->output.i_chroma );

    switch( p_vout->output.i_chroma )
    {
        /* I420 target, no scaling */
        case VLC_FOURCC('I','4','2','0'):
        case VLC_FOURCC('I','Y','U','V'):
        case VLC_FOURCC('Y','V','1','2'):
            BlendI420( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop );
            break;

        /* RGB 555 - scaled */
        case VLC_FOURCC('R','V','1','5'):
            BlendRV16( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop, 
                       VLC_TRUE );
            break;
          
        case VLC_FOURCC('R','V','1','6'):
            BlendRV16( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop,
                       /* Not sure under what conditions RV16 is really
                          RV16 and not RV15.
                        */
#if 0
                       VLC_FALSE );
#else
                       VLC_TRUE );
#endif
	    break;

        /* RV24 target, scaling */
        case VLC_FOURCC('R','V','2','4'):
            BlendRV24( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop );
	    break;

        /* RV32 target, scaling */
        case VLC_FOURCC('R','V','3','2'):
            BlendRV32( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop );
	    break;

        /* NVidia overlay, no scaling */
        case VLC_FOURCC('Y','U','Y','2'):
            BlendYUY2( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop );
	    break;

        /* Palettized 8 bits per pixel (256 colors). Each
           pixel is an uint8_t index in the palette
           Used in ASCII Art. 
        */
        case VLC_FOURCC('R','G','B','2'):
            BlendRGB2( p_vout, p_pic, p_spu, p_spu->p_sys->b_crop );
          
	    break;

        default:
            msg_Err( p_vout, "unknown chroma, can't render SPU" );
            break;
    }
}

/* Following functions are local */

/* 
  YV12 format:  

  All Y samples are found first in memory as an array of bytes
  (possibly with a larger stride for memory alignment), followed
  immediately by all Cr (=U) samples (with half the stride of the Y
  lines, and half the number of lines), then followed immediately by
  all Cb (=V) samples in a similar fashion.
*/

static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop )
{
  /* Common variables */
  uint8_t *p_pixel_base_Y, *p_pixel_base_V, *p_pixel_base_U;
  ogt_yuvt_t *p_source; /* This is the where the subtitle pixels come from */

  int i_x, i_y;
  vlc_bool_t even_scanline = VLC_FALSE;

  /* Crop-specific */
  int i_x_start, i_y_start, i_x_end, i_y_end;
  /* int i=0; */

  const struct subpicture_sys_t *p_sys = p_spu->p_sys;
  
  dbg_print( (DECODE_DBG_CALL|DECODE_DBG_RENDER), 
	     "spu width x height: (%dx%d), (x,y)=(%d,%d), yuv pitch (%d,%d,%d)", 
	     p_spu->i_width,  p_spu->i_height, p_spu->i_x, p_spu->i_y,
	     p_pic->Y_PITCH, p_pic->U_PITCH, p_pic->V_PITCH );

  
  p_pixel_base_Y = p_pic->p[Y_PLANE].p_pixels + p_spu->i_x
                 + p_pic->p[Y_PLANE].i_pitch * p_spu->i_y;
  
  p_pixel_base_U = p_pic->p[U_PLANE].p_pixels + p_spu->i_x/2
                 + p_pic->p[U_PLANE].i_pitch * p_spu->i_y/2;
  
  p_pixel_base_V = p_pic->p[V_PLANE].p_pixels + p_spu->i_x/2
                 + p_pic->p[V_PLANE].i_pitch * p_spu->i_y/2;
  
  i_x_start = p_sys->i_x_start;
  i_y_start = p_pic->p[Y_PLANE].i_pitch * p_sys->i_y_start;
  
  i_x_end   = p_sys->i_x_end;
  i_y_end   = p_pic->p[Y_PLANE].i_pitch * p_sys->i_y_end;
  
  p_source = (ogt_yuvt_t *)p_sys->p_data;
  
  /* Draw until we reach the bottom of the subtitle */
  for( i_y = 0; 
       i_y < p_spu->i_height * p_pic->p[Y_PLANE].i_pitch ;
       i_y += p_pic->p[Y_PLANE].i_pitch )
    {
      uint8_t *p_pixel_base_Y_y = p_pixel_base_Y + i_y;
      uint8_t *p_pixel_base_U_y = p_pixel_base_U + i_y/4;
      uint8_t *p_pixel_base_V_y = p_pixel_base_V + i_y/4;

      i_x = 0;

      if ( b_crop ) {
        if ( i_y > i_y_end ) break;
        if (i_x_start) {
          i_x = i_x_start;
          p_source += i_x_start;
        }
      }

      even_scanline = !even_scanline;

      /* Draw until we reach the end of the line */
      for( ; i_x < p_spu->i_width; i_x++, p_source++ )
	{

	  if( b_crop ) {

            /* FIXME: y cropping should be dealt with outside of this loop.*/
            if ( i_y < i_y_start) continue;

            if ( i_x > i_x_end )
	    {
	      p_source += p_spu->i_width - i_x;
              break;
	    }
          }

#ifdef TESTING_TRANSPARENCY
          if (p_source->s.t == MAX_TRANS) p_source->s.t >>= 1;
#endif

	  switch( p_source->s.t )
	    {
	    case 0: 
	      /* Completely transparent. Don't change pixel. */
	      break;
	      
	    case MAX_TRANS:
	      {
		/* Completely opaque. Completely overwrite underlying
		   pixel with subtitle pixel. */
		
		/* This is the location that's going to get changed.*/
		uint8_t *p_pixel_Y = p_pixel_base_Y_y + i_x;
		
		*p_pixel_Y = p_source->plane[Y_PLANE];

		if ( even_scanline && i_x % 2 == 0 ) {
		  uint8_t *p_pixel_U = p_pixel_base_U_y + i_x/2;
		  uint8_t *p_pixel_V = p_pixel_base_V_y + i_x/2;
		  *p_pixel_U = p_source->plane[U_PLANE];
		  *p_pixel_V = p_source->plane[V_PLANE];
		}
		
		break;
	      }
	      
	    default:
	      {
		/* Blend in underlying subtitle pixel. */
		
		/* This is the location that's going to get changed. */
		uint8_t *p_pixel_Y = p_pixel_base_Y_y + i_x;


		/* This is the weighted part of the subtitle. The
		   color plane is 8 bits and transparancy is 4 bits so
		   when multiplied we get up to 12 bits.
		 */
		uint16_t i_sub_color_Y = 
		  (uint16_t) ( p_source->plane[Y_PLANE] *
			       (uint16_t) (p_source->s.t) );

		/* This is the weighted part of the underlying pixel.
		   For the same reasons above, the result is up to 12
		   bits.  However since the transparancies are
		   inverses, the sum of i_sub_color and i_pixel_color
		   will not exceed 12 bits.
		*/
		uint16_t i_pixel_color_Y = 
		  (uint16_t) ( *p_pixel_Y * 
			       (uint16_t) (MAX_TRANS - p_source->s.t) ) ;
		
		/* Scale the 12-bit result back down to 8 bits. A
		   precise scaling after adding the two components,
		   would divide by one less than a power of 2. However
		   to simplify and speed things we use a power of
		   2. This means the boundaries (either all
		   transparent and all opaque) aren't handled properly.
		   But we deal with them in special cases above. */

		*p_pixel_Y = ( i_sub_color_Y + i_pixel_color_Y ) >> TRANS_BITS;

		if ( even_scanline && i_x % 2 == 0 ) {
		  uint8_t *p_pixel_U = p_pixel_base_U_y + i_x/2
                    - p_pic->p[U_PLANE].i_pitch / 2;
		  uint8_t *p_pixel_V = p_pixel_base_V_y + i_x/2
                    - p_pic->p[V_PLANE].i_pitch / 2;
		  uint16_t i_sub_color_U = 
		    (uint16_t) ( p_source->plane[U_PLANE] *
				 (uint16_t) (p_source->s.t) );
		  
		  uint16_t i_sub_color_V = 
		    (uint16_t) ( p_source->plane[V_PLANE] *
				 (uint16_t) (p_source->s.t) );
		  uint16_t i_pixel_color_U = 
		    (uint16_t) ( *p_pixel_U * 
				 (uint16_t) (MAX_TRANS - p_source->s.t) ) ;
		  uint16_t i_pixel_color_V = 
		    (uint16_t) ( *p_pixel_V * 
				 (uint16_t) (MAX_TRANS - p_source->s.t) ) ;
		  *p_pixel_U = ( i_sub_color_U + i_pixel_color_U )>>TRANS_BITS;
		  *p_pixel_V = ( i_sub_color_V + i_pixel_color_V )>>TRANS_BITS;
		}
		break;
	      }
	      
	    }
	}
    }
}

/*

  YUY2 Format:

  Data is found in memory as an array of bytes in which the first byte
  contains the first sample of Y, the second byte contains the first
  sample of Cb (=U), the third byte contains the second sample of Y,
  the fourth byte contains the first sample of Cr (=V); and so
  on. Each 32-bit word then contains information for two contiguous
  horizontal pixels, two 8-bit Y values plus a single Cb and Cr which
  spans the two pixels.
*/

#define BYTES_PER_PIXEL 4

static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
                        const subpicture_t *p_spu, vlc_bool_t b_crop )
{
  /* Common variables */
  uint8_t *p_pixel_base;

  /* This is the where the subtitle pixels come from */
  ogt_yuvt_t *p_source = (ogt_yuvt_t *) p_spu->p_sys->p_data;

  ogt_yuvt_t *p_source_end = (ogt_yuvt_t *)p_spu->p_sys->p_data +