Skip to content

V
vlc
Commit a11a1831 authored by Rocky Bernstein's avatar Rocky Bernstein
Browse files
Options

RGB2 transparency now works. Common 8-bit RGB averaging routine 

hoisted. Some doc additions.
parent b0e78a0a
Show whitespace changes
Inline Side-by-side
Showing with 182 additions and 144 deletions
modules/codec/ogt/render.c
View file @ a11a1831
/*****************************************************************************
* render.c : Philips OGT and CVD (VCD Subtitle) blending routines
* 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 1.27 2004/01/31 05:53:35 rocky Exp $
* $Id: render.c,v 1.28 2004/02/01 05:15:22 rocky Exp $
*
* Author: Rocky Bernstein <rocky@panix.com>
* based on code from:
......@@ -38,15 +40,16 @@
#include "subtitle.h"
#include "render.h"
/* We use 4 bits for an alpha value: 0..15, 15 is completely transparent and
0 completely opaque. Note that although SVCD allow 8-bits, pixels
previously should be scaled down to 4 bits to use these routines.
/* 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 ALPHA_BITS (4)
#define MAX_ALPHA ((1<<ALPHA_BITS) - 1)
#define ALPHA_SCALEDOWN (8-ALPHA_BITS)
#define TRANS_BITS (4)
#define MAX_TRANS ((1<<TRANS_BITS) - 1)
#define TRANS_SCALEDOWN (8-TRANS_BITS)
/* We use a fixed-point arithmetic to scaling ratios so that we
/* 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.
*/
......@@ -55,6 +58,30 @@
/* 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
*****************************************************************************/
......@@ -236,7 +263,7 @@ static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
}
#ifdef TESTING_TRANSPARENCY
if (p_source->s.t == MAX_ALPHA) p_source->s.t >>= 1;
if (p_source->s.t == MAX_TRANS) p_source->s.t >>= 1;
#endif
switch( p_source->s.t )
......@@ -245,7 +272,7 @@ static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -289,7 +316,7 @@ static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
*/
uint16_t i_pixel_color_Y =
(uint16_t) ( *p_pixel_Y *
(uint16_t) (MAX_ALPHA - p_source->s.t) ) ;
(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,
......@@ -299,7 +326,7 @@ static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
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 ) >> ALPHA_BITS;
*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
......@@ -315,12 +342,12 @@ static void BlendI420( vout_thread_t *p_vout, picture_t *p_pic,
(uint16_t) (p_source->s.t) );
uint16_t i_pixel_color_U =
(uint16_t) ( *p_pixel_U *
(uint16_t) (MAX_ALPHA - p_source->s.t) ) ;
(uint16_t) (MAX_TRANS - p_source->s.t) ) ;
uint16_t i_pixel_color_V =
(uint16_t) ( *p_pixel_V *
(uint16_t) (MAX_ALPHA - p_source->s.t) ) ;
*p_pixel_U = ( i_sub_color_U + i_pixel_color_U )>>ALPHA_BITS;
*p_pixel_V = ( i_sub_color_V + i_pixel_color_V )>>ALPHA_BITS;
(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;
}
......@@ -427,15 +454,15 @@ static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
/* Favor opaque subtitle pixels. */
if ( (p_source->s.t == 0) && (p_source+1)->s.t == MAX_ALPHA )
if ( (p_source->s.t == 0) && (p_source+1)->s.t == MAX_TRANS )
i_avg_tr = (p_source+1)->s.t;
else if ( (p_source->s.t == MAX_ALPHA) && (p_source+1)->s.t == 0 )
else if ( (p_source->s.t == MAX_TRANS) && (p_source+1)->s.t == 0 )
i_avg_tr = p_source->s.t;
else
i_avg_tr = ( p_source->s.t + (p_source+1)->s.t ) / 2;
#ifdef TESTING_TRANSPARENCY
if (i_avg_tr == MAX_ALPHA) i_avg_tr >>= 1;
if (i_avg_tr == MAX_TRANS) i_avg_tr >>= 1;
#endif
switch( i_avg_tr )
......@@ -444,7 +471,7 @@ static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -455,10 +482,10 @@ static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
uint8_t i_avg_v;
/* Favor opaque subtitle pixel. */
if (p_source->s.t == MAX_ALPHA ) {
if (p_source->s.t == MAX_TRANS ) {
i_avg_u = p_source->plane[U_PLANE] ;
i_avg_v = p_source->plane[V_PLANE] ;
} else if ( (p_source+1)->s.t == MAX_ALPHA ) {
} else if ( (p_source+1)->s.t == MAX_TRANS ) {
i_avg_u = (p_source+1)->plane[U_PLANE] ;
i_avg_v = (p_source+1)->plane[V_PLANE] ;
} else {
......@@ -514,16 +541,16 @@ static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
uint16_t i_pixel_color_Y1 =
(uint16_t) ( *(p_pixel) *
(uint16_t) (MAX_ALPHA - i_avg_tr) ) ;
(uint16_t) (MAX_TRANS - i_avg_tr) ) ;
uint16_t i_pixel_color_Y2 =
(uint16_t) ( *(p_pixel+2) *
(uint16_t) (MAX_ALPHA - i_avg_tr) ) ;
(uint16_t) (MAX_TRANS - i_avg_tr) ) ;
uint16_t i_pixel_color_U =
(uint16_t) ( *(p_pixel+1) *
(uint16_t) (MAX_ALPHA - i_avg_tr) ) ;
(uint16_t) (MAX_TRANS - i_avg_tr) ) ;
uint16_t i_pixel_color_V =
(uint16_t) ( *(p_pixel+3) *
(uint16_t) (MAX_ALPHA - i_avg_tr) ) ;
(uint16_t) (MAX_TRANS - i_avg_tr) ) ;
/* draw a two contiguous pixels: 2 Y values, 1 U, and 1 V. */
......@@ -535,10 +562,10 @@ static void BlendYUY2( vout_thread_t *p_vout, picture_t *p_pic,
transparent and all opaque) aren't handled properly.
But we deal with them in special cases above. */
*p_pixel++ = ( i_sub_color_Y1 + i_pixel_color_Y1 )>>ALPHA_BITS;
*p_pixel++ = ( i_sub_color_U + i_pixel_color_U ) >>ALPHA_BITS;
*p_pixel++ = ( i_sub_color_Y2 + i_pixel_color_Y2 )>>ALPHA_BITS;
*p_pixel++ = ( i_sub_color_V + i_pixel_color_V ) >>ALPHA_BITS;
*p_pixel++ = ( i_sub_color_Y1 + i_pixel_color_Y1 )>>TRANS_BITS;
*p_pixel++ = ( i_sub_color_U + i_pixel_color_U ) >>TRANS_BITS;
*p_pixel++ = ( i_sub_color_Y2 + i_pixel_color_Y2 )>>TRANS_BITS;
*p_pixel++ = ( i_sub_color_V + i_pixel_color_V ) >>TRANS_BITS;
break;
}
}
......@@ -623,68 +650,62 @@ yuv2rgb565(ogt_yuvt_t *p_yuv, uint8_t *p_rgb1, uint8_t *p_rgb2 )
static inline void
rv16_pack_blend(uint8_t *p_pixel, ogt_yuvt_t *p_source, uint8_t *p_rgb1,
uint8_t *p_rgb2, vlc_bool_t b_15bpp, uint8_t i_alpha,
uint8_t *p_rgb2, vlc_bool_t b_15bpp, uint8_t i_trans,
int a_scale_down )
{
uint8_t rgb_source[3];
uint8_t rgb[RGB_SIZE];
uint8_t rgb_sub[3];
uint8_t rgb_vout[RGB_SIZE];
int i;
#ifdef FIXED_RV16_TRANSPARENCY
uint8_t i_dest_alpha = MAX_ALPHA - i_alpha;
#endif
yuv2rgb(p_source, rgb_source);
yuv2rgb(p_source, rgb_sub);
/* Scale RGB from 8 bits down to 6 or 5. */
rgb_source[GREEN_PIXEL] >>= (8-5);
rgb_source[BLUE_PIXEL] >>= (8-5);
rgb_sub[GREEN_PIXEL] >>= (8-5);
rgb_sub[BLUE_PIXEL] >>= (8-5);
rgb[GREEN_PIXEL] = *(p_pixel+1) & 0x1f;
rgb_vout[GREEN_PIXEL] = *(p_pixel+1) & 0x1f;
if (b_15bpp) {
rgb_source[RED_PIXEL] >>= (8-5);
rgb[BLUE_PIXEL] = ((*p_pixel)>>2) & 0x1f;
rgb[RED_PIXEL] = ((*p_pixel & 0x03) << 3) | ((*(p_pixel+1)&0xe0)>>5);
rgb_sub[RED_PIXEL] >>= (8-5);
rgb_vout[BLUE_PIXEL] = ((*p_pixel)>>2) & 0x1f;
rgb_vout[RED_PIXEL] = ((*p_pixel & 0x03) << 3) | ((*(p_pixel+1)&0xe0)>>5);
} else {
rgb_source[RED_PIXEL] >>= (8-6);
rgb[BLUE_PIXEL] = ((*p_pixel)>>3) & 0x1f;
rgb[RED_PIXEL] = ((*p_pixel & 0x07) << 3) | ((*(p_pixel+1)&0xe0)>>5);
rgb_sub[RED_PIXEL] >>= (8-6);
rgb_vout[BLUE_PIXEL] = ((*p_pixel)>>3) & 0x1f;
rgb_vout[RED_PIXEL] = ((*p_pixel & 0x07) << 3) | ((*(p_pixel+1)&0xe0)>>5);
}
#if 0
printf("r,g,b=(%d,%d,%d), source r,g,b=(%d,%d,%d), alpha %d, dest_alpha %d\n",
rgb[RED_PIXEL], rgb[GREEN_PIXEL], rgb[BLUE_PIXEL],
rgb_source[RED_PIXEL], rgb_source[GREEN_PIXEL],
rgb_source[BLUE_PIXEL], i_alpha, i_dest_alpha);
printf("r,g,b=(%d,%d,%d), sub r,g,b=(%d,%d,%d), trans %d, inv_trans %d\n",
rgb_vout[RED_PIXEL], rgb_vout[GREEN_PIXEL], rgb_vout[BLUE_PIXEL],
rgb_sub[RED_PIXEL], rgb_sub[GREEN_PIXEL],
rgb_sub[BLUE_PIXEL], i_trans, i_inv_trans);
#endif
for (i=0; i < RGB_SIZE; i++) {
/* Average the two pixels. */
#ifdef FIXED_RV16_TRANSPARENCY
rgb[i] = ( (uint16_t) rgb_source[i]*i_alpha + rgb[i]*i_dest_alpha) >>
ALPHA_BITS;
avg_8bit_rgb(rgb_vout, rgb_sub, i_trans);
#else
for (i=0; i < RGB_SIZE; i++) {
/* For now the Best we can do is fade the color. Picking up
video underneath doesn't work. */
/* rgb[i] = ( (uint16_t) rgb[i]*i_dest_alpha ) >> ALPHA_BITS; */
rgb[i] = ( (uint16_t) rgb_source[i]*i_alpha ) >> ALPHA_BITS;
/* rgb_vout[i] = ( (uint16_t) rgb_vout[i]*i_inv_trans ) >> TRANS_BITS; */
rgb_vout[i] = ( (uint16_t) rgb_sub[i]*i_trans ) >> TRANS_BITS;
#endif
}
#if 0
printf("avg r,g,b=(%d,%d,%d)\n",
rgb[RED_PIXEL], rgb[GREEN_PIXEL], rgb[BLUE_PIXEL] );
rgb_vout[RED_PIXEL], rgb_vout[GREEN_PIXEL], rgb_vout[BLUE_PIXEL] );
#endif
#if 0
if (b_15bpp) {
*p_rgb1 = ( (rgb[BLUE_PIXEL] << 2)&0x7c )|( (rgb[RED_PIXEL]>>3)&0x03 );
*p_rgb1 = ( (rgb_vout[BLUE_PIXEL] << 2)&0x7c )|( (rgb_vout[RED_PIXEL]>>3)&0x03 );
} else {
*p_rgb1 = ( (rgb[BLUE_PIXEL] << 3)&0xF8 )|( (rgb[RED_PIXEL]>>3)&0x07 );
*p_rgb1 = ( (rgb_vout[BLUE_PIXEL] << 3)&0xF8 )|( (rgb_vout[RED_PIXEL]>>3)&0x07 );
}
*p_rgb2 = ( (rgb[RED_PIXEL] << 5)&0xe0 ) | ( rgb[GREEN_PIXEL]&0x1f );
*p_rgb2 = ( (rgb_vout[RED_PIXEL] << 5)&0xe0 ) | ( rgb_vout[GREEN_PIXEL]&0x1f );
#else
*p_rgb1 = (*p_rgb1)+1;
*p_rgb2 = (*p_rgb2)+1;
......@@ -802,7 +823,7 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
}
#ifdef TESTING_TRANSPARENCY
if (p_source->s.t == MAX_ALPHA) p_source->s.t >>= 1;
if (p_source->s.t == MAX_TRANS) p_source->s.t >>= 1;
#endif
switch( p_source->s.t )
......@@ -811,7 +832,7 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -857,7 +878,7 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
for ( len = i_xlast - i_xdest; len ; len--) {
rv16_pack_blend(p_dest, p_source, &i_rgb1, &i_rgb2,
b_15bpp, p_source->s.t, ALPHA_SCALEDOWN);
b_15bpp, p_source->s.t, TRANS_SCALEDOWN);
*p_dest++ = i_rgb1;
*p_dest++ = i_rgb2;
}
......@@ -900,7 +921,7 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -944,7 +965,7 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
uint8_t *p_dest = p_pixel_base + i_ytmp;
for ( len = i_xlast - i_xdest; len ; len--) {
rv16_pack_blend(p_dest, p_source, &i_rgb1, &i_rgb2,
b_15bpp, p_source->s.t, ALPHA_SCALEDOWN);
b_15bpp, p_source->s.t, TRANS_SCALEDOWN);
*p_dest++ = i_rgb1;
*p_dest++ = i_rgb2;
}
......@@ -961,33 +982,26 @@ BlendRV16( vout_thread_t *p_vout, picture_t *p_pic,
#define BYTES_PER_PIXEL 4
static inline void
rv24_pack_blend(uint8_t *p_pixel, uint8_t *rgb_source, uint8_t i_alpha,
rv24_pack_blend(uint8_t *rgb_vout, const uint8_t *rgb_sub, uint8_t i_trans,
int a_scale_down )
{
int i;
uint8_t i_dest_alpha = MAX_ALPHA - i_alpha;
#if 0
printf("r,g,b=(%d,%d,%d), source r,g,b=(%d,%d,%d), alpha %d, dest_alpha %d\n",
p_pixel[RED_PIXEL], p_pixel[GREEN_PIXEL], p_pixel[BLUE_PIXEL],
rgb_source[RED_PIXEL], rgb_source[GREEN_PIXEL],
rgb_source[BLUE_PIXEL], i_alpha, i_dest_alpha);
printf("r,g,b=(%d,%d,%d), source r,g,b=(%d,%d,%d), trans %d\n",
rgb_vout[RED_PIXEL], rgb_vout[GREEN_PIXEL], rgb_vout[BLUE_PIXEL],
rgb_sub[RED_PIXEL], rgb_sub[GREEN_PIXEL],
rgb_sub[BLUE_PIXEL], i_trans);
#endif
#ifdef WORDS_BIGENDIAN
*p_pixel++;
*rgb_vout++;
#endif
for (i=0; i < RGB_SIZE; i++) {
/* Average the two pixels. */
p_pixel[i] = ( (uint16_t) rgb_source[i]*i_alpha + p_pixel[i]*i_dest_alpha)
>> ALPHA_BITS;
}
avg_8bit_rgb(rgb_vout, rgb_sub, i_trans);
#if 0
printf("avg r,g,b=(%d,%d,%d)\n",
rgb[RED_PIXEL], rgb[GREEN_PIXEL], rgb[BLUE_PIXEL] );
rgb_vout[RED_PIXEL], rgb_vout[GREEN_PIXEL], rgb_vout[BLUE_PIXEL] );
#endif
}
......@@ -1111,7 +1125,7 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
}
#ifdef TESTING_TRANSPARENCY
if (p_source->s.t == MAX_ALPHA) p_source->s.t >>= 2;
if (p_source->s.t == MAX_TRANS) p_source->s.t >>= 2;
#endif
switch( p_source->s.t )
......@@ -1120,7 +1134,7 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -1153,10 +1167,12 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
* BYTES_PER_PIXEL );
uint32_t len = i_xlast - i_xdest * BYTES_PER_PIXEL;
/* To be able to scale correctly for full opaqueness, we
add 1 to the alpha. This means alpha value 0 won't
be completely transparent and is not correct, but
that's handled in a special case above anyway. */
/* To be able to scale correctly for full
opaqueness, we add 1 to the transparency.
This means transparency value 0 won't be
completely transparent which is not correct.
But that's handled in a special case above
anyway. */
/* This is the location that's going to get changed. */
uint8_t *p_dest = p_pixel_base_y + i_xdest;
......@@ -1166,7 +1182,7 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
for ( len = i_xlast - i_xdest; len ; len--) {
rv24_pack_blend(p_dest, rgb, p_source->s.t,
ALPHA_SCALEDOWN);
TRANS_SCALEDOWN);
p_dest += BYTES_PER_PIXEL;
}
break;
......@@ -1209,7 +1225,7 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
/* Completely transparent. Don't change pixel. */
break;
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -1267,7 +1283,7 @@ BlendRV24( vout_thread_t *p_vout, picture_t *p_pic,
for ( len = i_xlast - i_xdest; len ; len--) {
rv24_pack_blend(p_dest, rgb, p_source->s.t,
ALPHA_SCALEDOWN);
TRANS_SCALEDOWN);
p_dest += BYTES_PER_PIXEL;
}
}
......@@ -1407,7 +1423,7 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
break;
default:
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -1448,7 +1464,7 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
be completely transparent and is not correct, but
that's handled in a special case above anyway. */
uint8_t i_destalpha = MAX_ALPHA - p_source->s.t;
uint8_t i_destalpha = MAX_TRANS - p_source->s.t;
uint8_t rgb[RGB_SIZE];
/* This is the location that's going to get changed. */
......@@ -1456,7 +1472,7 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
yuv2rgb(p_source, rgb);
rv32_pack_blend(p_dest, rgb, dest_alpha,
ALPHA_SCALEDOWN);
TRANS_SCALEDOWN);
for ( len = i_xlast - i_xdest; len ; len--) {
*p_dest++ = rgb[BLUE_PIXEL];
......@@ -1506,7 +1522,7 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
break;
default:
case MAX_ALPHA:
case MAX_TRANS:
{
/* Completely opaque. Completely overwrite underlying
pixel with subtitle pixel. */
......@@ -1562,9 +1578,9 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
be completely transparent and is not correct, but
that's handled in a special case above anyway. */
uint8_t i_destalpha = MAX_ALPHA - p_source->s.t;
uint8_t i_destalpha = MAX_TRANS - p_source->s.t;
rv32_pack_blend(p_dest, rgb, dest_alpha,
ALPHA_SCALEDOWN);
TRANS_SCALEDOWN);
}
break;
#endif /*TRANSPARENCY_FINISHED*/
......@@ -1574,60 +1590,68 @@ BlendRV32( vout_thread_t *p_vout, picture_t *p_pic,
}
}
/* 4-entry array of colormap indices */
static uint8_t cmap[NUM_SUBTITLE_COLORS];
/* Actual RGB values for above; this is used in blending.*/
static uint8_t cmap_rgb[NUM_SUBTITLE_COLORS][RGB_SIZE];
/*
Return the colormap index for the average of p_pixel and a subtitle
pixel in RGB form.
Return the colormap index for the average of i_pixel and a subtitle
pixel whose subtitle palette entry is i_cmap.
*/
static inline cmap_t
avg_rgb2(const vout_thread_t *p_vout, uint8_t i_pixel, cmap_t i_cmap_sub,
const uint8_t rgb_sub[] )
avg_rgb2(const vout_thread_t *p_vout, uint8_t i_pixel, uint8_t i_trans,
int i_cmap, mtime_t i_pts)
{
uint8_t rgb_vout[RGB_SIZE];
static cmap_t avg_cache[CMAP_RGB2_SIZE][NUM_SUBTITLE_COLORS];
static vlc_bool_t b_first_time = VLC_TRUE;
int i;
/* FIXME: really we need to save subtitle number since in theory
the palette can change each on each distinct subtitle. In practice
this doesn't happen that much.
/* Cache the average of a vout colormap entry and a subtitle palette
entry. There are not too many of these 256*4 = 1024.
*/
if (b_first_time)
static cmap_t avg_cache[CMAP_RGB2_SIZE][NUM_SUBTITLE_COLORS];
/* subtitle palettes might change between two subtitles. i_last_pts
will be used to figure out if the subtitle has changed, and
we have to invalidate the cache. */
static mtime_t i_last_pts = -1;
if (i_pts != i_last_pts)
{
int i, j;
for (i=0; i<CMAP_RGB2_SIZE; i++)
for (j=0; j<NUM_SUBTITLE_COLORS; j++)
avg_cache[i][j] = INVALID_CMAP_ENTRY;
/* Hack: We rely on the fact that INVALID_CMAP_ENTRY is repeated
0xFF.
*/
memset(avg_cache, 0xFF, sizeof(avg_cache));
i_last_pts = i_pts;
}
if ( avg_cache[i_pixel][i_cmap_sub] != INVALID_CMAP_ENTRY )
return avg_cache[i_pixel][i_cmap_sub];
if ( avg_cache[i_pixel][i_cmap] != INVALID_CMAP_ENTRY )
return avg_cache[i_pixel][i_cmap];
if ( !query_color(p_vout, i_pixel, rgb_vout) ) return INVALID_CMAP_ENTRY;
for (i = 0; i < RGB_SIZE; i++)
{
rgb_vout[i] = (rgb_vout[i] + rgb_sub[i]) / 2;
}
avg_8bit_rgb(rgb_vout, cmap_rgb[i_cmap], i_trans);
#if 0
{
uint8_t rgb_approx[RGB_SIZE];
avg_cache[i_pixel][i_cmap_sub] =
avg_cache[i_pixel][i_cmap] =
find_cmap_rgb8_nearest(p_vout, rgb_vout, rgb_approx);
printf(
"cmap old %0x new 0%x sub=(%0x, %0x, %0x) "
"avg=(%0x, %0x, %0x) vout=(%0x, %0x, %0x)\n",
i_pixel, i_cmap,
rgb_sub[RED_PIXEL], rgb_sub[GREEN_PIXEL], rgb_sub[BLUE_PIXEL],
"cmap old %0x avg approx 0%x sub: %d sub=(%0x, %0x, %0x) "
"approx=(%0x, %0x, %0x) avg vout=(%0x, %0x, %0x)\n",
i_pixel, avg_cache[i_pixel][i_cmap], i_cmap,
cmap_rgb[i_cmap][RED_PIXEL], cmap_rgb[i_cmap][GREEN_PIXEL],
cmap_rgb[i_cmap][BLUE_PIXEL],
rgb_approx[RED_PIXEL], rgb_approx[GREEN_PIXEL], rgb_approx[BLUE_PIXEL],
rgb_vout[RED_PIXEL], rgb_vout[GREEN_PIXEL], rgb_vout[BLUE_PIXEL]);
}
#else
avg_cache[i_pixel][i_cmap_sub] =
avg_cache[i_pixel][i_cmap] =
find_cmap_rgb8_nearest(p_vout, rgb_vout, NULL);
#endif
return avg_cache[i_pixel][i_cmap_sub];
return avg_cache[i_pixel][i_cmap];
}
#undef BYTES_PER_PIXEL
......@@ -1642,6 +1666,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
uint8_t *p_src_start = (uint8_t *)p_spu->p_sys->p_data;
uint8_t *p_src_end = &p_src_start[p_spu->i_height * p_spu->i_width];
uint8_t *p_source; /* This is the where the subtitle pixels come from */
int i;
int i_x, i_y;
int i_y_src;
......@@ -1657,13 +1682,6 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
/* Crop-specific */
int i_x_start, i_y_start, i_x_end, i_y_end;
/* 4-entry array of colormap indices */
uint8_t cmap[NUM_SUBTITLE_COLORS];
int i;
/* Actual RGB values for above; this is used in blending.*/
uint8_t cmap_rgb[NUM_SUBTITLE_COLORS][RGB_SIZE];
struct subpicture_sys_t *p_sys = p_spu->p_sys;
unsigned int i_aspect_x, i_aspect_y;
......@@ -1694,10 +1712,9 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
if ( p_sys->p_palette[i].s.t != 0 ) {
uint8_t rgb[RGB_SIZE];
uint8_t approx_rgb[RGB_SIZE];
yuv2rgb(&(p_sys->p_palette[i]), rgb);
cmap[i] =
find_cmap_rgb8_nearest(p_vout, rgb, approx_rgb);
find_cmap_rgb8_nearest(p_vout, rgb, cmap_rgb[i]);
dbg_print( (DECODE_DBG_RENDER),
"palette %d RGB=(%0x, %0x, %0x)\n", i,
rgb[RED_PIXEL], rgb[GREEN_PIXEL], rgb[BLUE_PIXEL]);
......@@ -1768,7 +1785,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
p_yuvt = p_sys->p_palette[*p_source & 0x3];
#ifdef TESTING_TRANSPARENCY
if (p_yuvt.s.t == MAX_ALPHA) p_yuvt.s.t >>= 1;
if (p_yuvt.s.t == MAX_TRANS) p_yuvt.s.t >>= 1;
#endif
switch ( p_yuvt.s.t )
......@@ -1779,7 +1796,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
printf(" "); /*++++*/
#endif
break;
case MAX_ALPHA:
case MAX_TRANS:
{
uint32_t i_xdest = ( ((i_x*i_xscale) >> ASCALE)
* BYTES_PER_PIXEL );
......@@ -1803,15 +1820,25 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
uint8_t *p_pixel = p_pixel_base_y + i_xdest;
uint32_t len = i_xlast - i_xdest;
#if FULL_TRANSPARENCY
/* This is what should be done, but it may be too
slow. */
for ( len = i_xlast - i_xdest -1; len >= 0; len-- )
{
cmap_t i_cmap = avg_rgb2(p_vout, *p_pixel,
cmap[*p_source],
cmap_rgb[*p_source]);
cmap_t i_cmap = avg_rgb2(p_vout, *p_pixel, p_yuvt.s.t,
*p_source, p_sys->i_pts);
if (i_cmap != INVALID_CMAP_ENTRY)
*p_pixel= (uint8_t) i_cmap;
p_pixel++;
}
#else
cmap_t i_cmap = avg_rgb2(p_vout, *p_pixel, p_yuvt.s.t,
*p_source, p_sys->i_pts);
if (i_cmap != INVALID_CMAP_ENTRY)
memset(p_pixel, i_cmap, len);
#endif
#if 0
printf("%1d", *p_source); /*++++*/
#endif
......@@ -1851,7 +1878,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
}
#ifdef TESTING_TRANSPARENCY
if (p_yuvt.s.t == MAX_ALPHA) p_yuvt.s.t >>= 1;
if (p_yuvt.s.t == MAX_TRANS) p_yuvt.s.t >>= 1;
#endif
switch ( p_yuvt.s.t )
{
......@@ -1861,7 +1888,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
printf(" "); /*++++*/
#endif
break;
case MAX_ALPHA:
case MAX_TRANS:
{
uint32_t i_xdest = ( ((i_x*i_xscale) >> ASCALE)
* BYTES_PER_PIXEL );
......@@ -1884,7 +1911,7 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
* BYTES_PER_PIXEL );
uint32_t i_xlast = ( (((i_x+1)*i_xscale) >> ASCALE)
* BYTES_PER_PIXEL );
int32_t len;
int32_t len = i_xlast - i_xdest;
#if 0
printf("%1d", *p_source); /*++++*/
#endif
......@@ -1893,15 +1920,26 @@ BlendRGB2( vout_thread_t *p_vout, picture_t *p_pic,
i_ytmp += p_pic->p->i_pitch ) {
/* This is the pixel that's going to change;*/
uint8_t *p_pixel = p_pixel_base + i_ytmp + i_xdest;
#if FULL_TRANSPARENCY
/* This is what should be done, but it may be too
slow. */
for ( len = i_xlast - i_xdest -1; len >= 0; len-- )
{
cmap_t i_cmap = avg_rgb2(p_vout, *p_pixel,
cmap[*p_source],
cmap_rgb[*p_source]);
p_yuvt.s.t, *p_source,
p_sys->i_pts);
if (i_cmap != INVALID_CMAP_ENTRY)
*p_pixel= (uint8_t) i_cmap;
p_pixel++;
}
#else
cmap_t i_cmap = avg_rgb2(p_vout, *p_pixel, p_yuvt.s.t,
*p_source, p_sys->i_pts);
if (i_cmap != INVALID_CMAP_ENTRY)
memset(p_pixel, i_cmap, len);
#endif
}
}
}
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment