diff options
Diffstat (limited to 'src/rmd_yuv_utils.c')
-rw-r--r-- | src/rmd_yuv_utils.c | 517 |
1 files changed, 517 insertions, 0 deletions
diff --git a/src/rmd_yuv_utils.c b/src/rmd_yuv_utils.c new file mode 100644 index 0000000..1f163ae --- /dev/null +++ b/src/rmd_yuv_utils.c @@ -0,0 +1,517 @@ +/****************************************************************************** +* recordMyDesktop - rmd_yuv_utils.c * +******************************************************************************* +* * +* Copyright (C) 2006,2007,2008 John Varouhakis * +* Copyright (C) 2008 Luca Bonavita * +* * +* 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 02111-1307 USA * +* * +* * +* * +* For further information contact me at johnvarouhakis@gmail.com * +******************************************************************************/ + +#include "config.h" +#include "rmd_yuv_utils.h" + +#include "rmd_math.h" + + +// Keep these global (for performance reasons I assume). +unsigned char Yr[256], Yg[256], Yb[256], + Ur[256], Ug[256], UbVr[256], + Vg[256], Vb[256]; + +// FIXME: These globals are modified in other source files! We keep +// thsee here for now. These are the cache blocks. They need to be +// accesible in the dbuf macros +unsigned char *yblocks, + *ublocks, + *vblocks; + +void rmdMakeMatrices (void) { + int i; + + /* assuming 8-bit precision */ + float Yscale = 219.0, Yoffset = 16.0; + float Cscale = 224.0, Coffset = 128.0; + float RGBscale = 255.0; + + float r, g, b; + float yr, yg, yb; + float ur, ug, ub; + float vg, vb; /* vr intentionally missing */ + + /* as for ITU-R BT-601-6 specifications: */ + r = 0.299; + b = 0.114; + g = 1.0 - r - b; + + /* as a note, here are the coefficients + as for ITU-R BT-709 specifications: + r=0.2126; b=0.0722; g=1.0-r-b; */ + + yr = r * Yscale / RGBscale; + yg = g * Yscale / RGBscale; + yb = b * Yscale / RGBscale; + ur = ( -0.5 * r / ( 1 - b ) ) * Cscale / RGBscale; + ug = ( -0.5 * g / ( 1 - b ) ) * Cscale / RGBscale; + ub = ( 0.5 * Cscale / RGBscale); + /* vr = ub so UbVr = ub*i = vr*i */ + vg = ( -0.5 * g / ( 1 - r ) ) * Cscale / RGBscale; + vb = ( -0.5 * b / ( 1 - r ) ) * Cscale / RGBscale; + + for (i = 0; i < 256; i++) { + Yr[i] = (unsigned char) rmdRoundf( Yoffset + yr * i ); + Yg[i] = (unsigned char) rmdRoundf( yg * i ); + Yb[i] = (unsigned char) rmdRoundf( yb * i ); + + Ur[i] = (unsigned char) rmdRoundf( Coffset + ur * i ); + Ug[i] = (unsigned char) rmdRoundf( ug * i ); + UbVr[i] = (unsigned char) rmdRoundf( ub * i ); + + Vg[i] = (unsigned char) rmdRoundf( vg * i ); + Vb[i] = (unsigned char) rmdRoundf( Coffset + vb * i ); + } +} + +static inline int blocknum(int xv, int yv, int widthv, int blocksize) +{ + return ((yv/blocksize) * (widthv/blocksize) + (xv/blocksize)); +} + +/* These at least make some sense as macros since they need duplication for + * the multiple depths, so I've just moved and reformatted them here for now. + */ + +#define UPDATE_Y_PLANE( data, \ + x_tm, \ + y_tm, \ + width_tm, \ + height_tm, \ + yuv, \ + __depth__) { \ + \ + register unsigned char *yuv_Y = (yuv)->y + x_tm + y_tm * (yuv)->y_stride, \ + *_yr = Yr, *_yg = Yg, *_yb = Yb; \ + register u_int##__depth__##_t *datapi = (u_int##__depth__##_t *)data; \ + \ + for (int k = 0; k < height_tm; k++) { \ + for (int i = 0; i < width_tm; i++) { \ + register u_int##__depth__##_t t_val = *datapi; \ + \ + *yuv_Y = _yr[__RVALUE_##__depth__(t_val)] + \ + _yg[__GVALUE_##__depth__(t_val)] + \ + _yb[__BVALUE_##__depth__(t_val)]; \ + datapi++; \ + yuv_Y++; \ + } \ + \ + yuv_Y += (yuv)->y_stride - width_tm; \ + } \ +} + +//when adding the r values, we go beyond +//the (16 bit)range of the t_val variable, but we are performing +//32 bit arithmetics, so there's no problem. +//(This note is useless, I'm just adding because +//the addition of the A components in CALC_TVAL_AVG_32, +//now removed as uneeded, produced an overflow which would have caused +//color distrtion, where it one of the R,G or B components) +#define CALC_TVAL_AVG_16(t_val, datapi, datapi_next) { \ + register u_int16_t t1, t2, t3, t4; \ + \ + t1 = *datapi; \ + t2 = *(datapi + 1); \ + t3 = *datapi_next; \ + t4 = *(datapi_next + 1); \ + \ + t_val = ((((t1 & __R16_MASK) + (t2 & __R16_MASK) + \ + (t3 & __R16_MASK) + (t4 & __R16_MASK)) / 4) & __R16_MASK) + \ + ((((t1 & __G16_MASK) + (t2 & __G16_MASK)+ \ + (t3 & __G16_MASK) + (t4 & __G16_MASK)) / 4) & __G16_MASK) + \ + ((((t1 & __B16_MASK) + (t2 & __B16_MASK) + \ + (t3 & __B16_MASK) + (t4 & __B16_MASK)) / 4) & __B16_MASK); \ +} + +//the 4 most significant bytes represent the A component which +//does not need to be added on t_val, as it is always unused +#define CALC_TVAL_AVG_32(t_val, datapi, datapi_next) { \ + register unsigned int t1, t2, t3, t4; \ + \ + t1 = *datapi; \ + t2 = *(datapi + 1); \ + t3 = *datapi_next; \ + t4 = *(datapi_next + 1); \ + \ + t_val = ((((t1 & 0x00ff0000) + (t2 & 0x00ff0000) + \ + (t3 & 0x00ff0000) + (t4 & 0x00ff0000)) / 4) & 0x00ff0000) + \ + ((((t1 & 0x0000ff00) + (t2 & 0x0000ff00) + \ + (t3 & 0x0000ff00) + (t4 & 0x0000ff00)) / 4) & 0x0000ff00) + \ + ((((t1 & 0x000000ff) + (t2 & 0x000000ff) + \ + (t3 & 0x000000ff) + (t4 & 0x000000ff)) / 4) & 0x000000ff); \ +} + +#define UPDATE_A_UV_PIXEL( yuv_U, \ + yuv_V, \ + t_val, \ + datapi, \ + datapi_next, \ + _ur,_ug,_ubvr,_vg,_vb, \ + sampling, \ + __depth__) \ + \ + if (sampling == __PXL_AVERAGE) { \ + CALC_TVAL_AVG_##__depth__(t_val, datapi, datapi_next) \ + } else \ + t_val = *(datapi); \ + \ + *(yuv_U) = _ur[__RVALUE_##__depth__(t_val)] + \ + _ug[__GVALUE_##__depth__(t_val)] + \ + _ubvr[__BVALUE_##__depth__(t_val)]; \ + \ + *(yuv_V) = _ubvr[__RVALUE_##__depth__(t_val)] + \ + _vg[__GVALUE_##__depth__(t_val)] + \ + _vb[__BVALUE_##__depth__(t_val)]; + +#define UPDATE_UV_PLANES( data, \ + x_tm, \ + y_tm, \ + width_tm, \ + height_tm, \ + yuv, \ + sampling, \ + __depth__) { \ + \ + register u_int##__depth__##_t t_val; \ + register unsigned char *yuv_U = (yuv)->u + x_tm / 2 + \ + (y_tm * (yuv)->uv_width) / 2, \ + *yuv_V = (yuv)->v + x_tm / 2 + \ + (y_tm * (yuv)->uv_width) / 2, \ + *_ur = Ur, *_ug = Ug, *_ubvr = UbVr, \ + *_vg = Vg, *_vb = Vb; \ + register u_int##__depth__##_t *datapi = (u_int##__depth__##_t *)data, \ + *datapi_next = NULL; \ + int w_odd = width_tm % 2, h_odd = height_tm % 2; \ + \ + if (sampling == __PXL_AVERAGE) \ + datapi_next = datapi + width_tm; \ + \ + for (int k = 0; k < height_tm - h_odd; k += 2) { \ + for (int i = 0; i < width_tm - w_odd; i += 2) { \ + UPDATE_A_UV_PIXEL( yuv_U, \ + yuv_V, \ + t_val, \ + datapi, \ + datapi_next, \ + _ur, _ug, _ubvr, _vg, _vb, \ + sampling, \ + __depth__); \ + \ + datapi += 2; \ + if (sampling == __PXL_AVERAGE) \ + datapi_next += 2; \ + yuv_U++; \ + yuv_V++; \ + } \ + \ + yuv_U += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + yuv_V += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + \ + datapi += width_tm + w_odd; \ + if (sampling == __PXL_AVERAGE) \ + datapi_next += width_tm + w_odd; \ + } \ +} + +#define UPDATE_Y_PLANE_DBUF( data, \ + data_back, \ + x_tm, \ + y_tm, \ + width_tm, \ + height_tm, \ + yuv, \ + __depth__) { \ + \ + register u_int##__depth__##_t t_val; \ + register unsigned char *yuv_Y = (yuv)->y + x_tm + y_tm * (yuv)->y_stride, \ + *_yr = Yr, *_yg = Yg, *_yb = Yb; \ + register u_int##__depth__##_t *datapi = (u_int##__depth__##_t *)data, \ + *datapi_back = (u_int##__depth__##_t *)data_back; \ + \ + for (int k = 0; k < height_tm; k++) { \ + for (int i = 0; i < width_tm; i++) { \ + if (*datapi != *datapi_back) { \ + t_val = *datapi; \ + *yuv_Y = _yr[__RVALUE_##__depth__(t_val)] + \ + _yg[__GVALUE_##__depth__(t_val)] + \ + _yb[__BVALUE_##__depth__(t_val)]; \ + yblocks[blocknum(x_tm + i, y_tm + k, (yuv)->y_width, Y_UNIT_WIDTH)] = 1;\ + } \ + datapi++; \ + datapi_back++; \ + yuv_Y++; \ + } \ + yuv_Y += (yuv)->y_stride - width_tm; \ + } \ +} + +#define UPDATE_UV_PLANES_DBUF( data, \ + data_back, \ + x_tm, \ + y_tm, \ + width_tm, \ + height_tm, \ + yuv, \ + sampling, \ + __depth__) { \ + \ + register u_int##__depth__##_t t_val; \ + register unsigned char *yuv_U = (yuv)->u + x_tm / 2 + \ + (y_tm * (yuv)->uv_width) / 2, \ + *yuv_V = (yuv)->v + x_tm / 2 + \ + (y_tm * (yuv)->uv_width) / 2, \ + *_ur = Ur, *_ug = Ug, *_ubvr = UbVr, \ + *_vg = Vg, *_vb = Vb; \ + \ + register u_int##__depth__##_t *datapi = (u_int##__depth__##_t *)data, \ + *datapi_next = NULL, \ + *datapi_back = (u_int##__depth__##_t *)data_back, \ + *datapi_back_next = NULL; \ + int w_odd = width_tm % 2, h_odd = height_tm % 2; \ + \ + if (sampling == __PXL_AVERAGE) { \ + datapi_next = datapi + width_tm; \ + datapi_back_next = datapi_back + width_tm; \ + \ + for (int k = 0; k < height_tm - h_odd; k += 2) { \ + for (int i = 0; i < width_tm - w_odd; i += 2) { \ + if ( (*datapi != *datapi_back || \ + (*(datapi + 1) != *(datapi_back + 1)) || \ + (*datapi_next != *datapi_back_next) || \ + (*(datapi_next + 1) != *(datapi_back_next + 1)))) { \ + \ + UPDATE_A_UV_PIXEL( yuv_U, \ + yuv_V, \ + t_val, \ + datapi, \ + datapi_next, \ + _ur,_ug,_ubvr,_vg,_vb, \ + sampling, \ + __depth__); \ + \ + ublocks[blocknum(x_tm + i, y_tm + k, (yuv)->y_width, Y_UNIT_WIDTH)] = 1; \ + vblocks[blocknum(x_tm + i, y_tm + k, (yuv)->y_width, Y_UNIT_WIDTH)] = 1; \ + } \ + \ + datapi += 2; \ + datapi_back += 2; \ + datapi_next += 2; \ + datapi_back_next += 2; \ + \ + yuv_U++; \ + yuv_V++; \ + } \ + \ + yuv_U += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + yuv_V += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + \ + datapi += width_tm + w_odd; \ + datapi_back += width_tm + w_odd; \ + datapi_next += width_tm + w_odd; \ + datapi_back_next += width_tm + w_odd; \ + } \ + } else { \ + for (int k = 0; k < height_tm - h_odd; k += 2) { \ + for (int i = 0; i < width_tm - w_odd; i += 2) { \ + if ((*datapi != *datapi_back)) { \ + UPDATE_A_UV_PIXEL( yuv_U, \ + yuv_V, \ + t_val, \ + datapi, \ + datapi_next, \ + _ur, _ug, _ubvr, _vg, _vb, \ + sampling, \ + __depth__); \ + \ + ublocks[blocknum(x_tm + i, y_tm + k, (yuv)->y_width, Y_UNIT_WIDTH)] = 1; \ + vblocks[blocknum(x_tm + i, y_tm + k, (yuv)->y_width, Y_UNIT_WIDTH)] = 1; \ + } \ + \ + datapi += 2; \ + datapi_back += 2; \ + \ + yuv_U++; \ + yuv_V++; \ + } \ + \ + yuv_U += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + yuv_V += ((yuv)->y_stride - (width_tm - w_odd * 2)) >> 1; \ + \ + datapi += width_tm + w_odd; \ + datapi_back += width_tm + w_odd; \ + } \ + } \ +} + +void rmdUpdateYuvBuffer( yuv_buffer *yuv, + unsigned char *data, + unsigned char *data_back, + int x_tm, + int y_tm, + int width_tm, + int height_tm, + int sampling, + int depth) { + + if (data_back == NULL) { + switch (depth) { + case 24: + case 32: + UPDATE_Y_PLANE(data, x_tm, y_tm, width_tm, height_tm, yuv, 32); + UPDATE_UV_PLANES(data, x_tm, y_tm, width_tm, height_tm, yuv, sampling, 32); + break; + case 16: + UPDATE_Y_PLANE(data, x_tm, y_tm, width_tm, height_tm, yuv, 16); + UPDATE_UV_PLANES(data, x_tm, y_tm, width_tm, height_tm, yuv, sampling, 16); + break; + default: + assert(0); + } + } else { + switch (depth) { + case 24: + case 32: + UPDATE_Y_PLANE_DBUF(data, data_back, x_tm, y_tm, width_tm, height_tm, yuv, 32); + UPDATE_UV_PLANES_DBUF(data, data_back, x_tm, y_tm, width_tm, height_tm, yuv, sampling, 32); + break; + case 16: + UPDATE_Y_PLANE_DBUF(data, data_back, x_tm, y_tm, width_tm, height_tm, yuv, 16); + UPDATE_UV_PLANES_DBUF(data, data_back, x_tm, y_tm, width_tm, height_tm, yuv, sampling, 16); + break; + default: + assert(0); + } + } +} + +void rmdDummyPointerToYuv( yuv_buffer *yuv, + unsigned char *data_tm, + int x_tm, + int y_tm, + int width_tm, + int height_tm, + int x_offset, + int y_offset, + unsigned char no_pixel) { + + int i, k, j = 0; + int x_2 = x_tm / 2, y_2 = y_tm / 2, y_width_2 = yuv->y_width/2; + + for (k = y_offset; k < y_offset + height_tm; k++) { + for (i = x_offset; i < x_offset + width_tm; i++) { + j = k * 16 + i; + + if (data_tm[j * 4] != no_pixel) { + yuv->y[x_tm + (i - x_offset) + ((k - y_offset) + y_tm) * yuv->y_width] = + Yr[data_tm[j * 4 + __RBYTE]] + + Yg[data_tm[j * 4 + __GBYTE]] + + Yb[data_tm[j * 4 + __BBYTE]]; + + if ((k % 2) && (i % 2)) { + yuv->u[x_2 + (i - x_offset) / 2 + ((k - y_offset) / 2 + y_2) * y_width_2] = + Ur[data_tm[(k * width_tm + i) * 4 + __RBYTE]] + + Ug[data_tm[(k * width_tm + i) * 4 + __GBYTE]] + + UbVr[data_tm[(k * width_tm + i) * 4 + __BBYTE]]; + yuv->v[x_2 + (i - x_offset) / 2 + ((k - y_offset) / 2 + y_2) * y_width_2] = + UbVr[data_tm[(k * width_tm + i) * 4 + __RBYTE]] + + Vg[data_tm[(k * width_tm + i) * 4 + __GBYTE]] + + Vb[data_tm[(k * width_tm + i) * 4 + __BBYTE]] ; + } + } + } + } + +} + +static inline unsigned char avg_4_pixels( unsigned char *data_array, + int width_img, + int k_tm, + int i_tm, + int offset) +{ + + return ((data_array[(k_tm*width_img+i_tm)*RMD_ULONG_SIZE_T+offset]+ + data_array[((k_tm-1)*width_img+i_tm)*RMD_ULONG_SIZE_T+offset]+ + data_array[(k_tm*width_img+i_tm-1)*RMD_ULONG_SIZE_T+offset]+ + data_array[((k_tm-1)*width_img+i_tm-1)*RMD_ULONG_SIZE_T+offset])/4); +} + +void rmdXFixesPointerToYuv( yuv_buffer *yuv, + unsigned char *data_tm, + int x_tm, + int y_tm, + int width_tm, + int height_tm, + int x_offset, + int y_offset, + int column_discard_stride) { + + unsigned char avg0, avg1, avg2, avg3; + int x_2 = x_tm / 2, y_2 = y_tm / 2; + + for (int k = y_offset; k < y_offset + height_tm; k++) { + for (int i = x_offset;i < x_offset + width_tm; i++) { + int j = k * (width_tm + column_discard_stride) + i; + + yuv->y[x_tm + (i - x_offset) + (k + y_tm - y_offset) * yuv->y_width] = + (yuv->y[x_tm + (i - x_offset) + (k - y_offset + y_tm) * yuv->y_width] * + (UCHAR_MAX - data_tm[(j * RMD_ULONG_SIZE_T) + __ABYTE]) + + ( ( Yr[data_tm[(j * RMD_ULONG_SIZE_T) + __RBYTE]] + + Yg[data_tm[(j * RMD_ULONG_SIZE_T) + __GBYTE]] + + Yb[data_tm[(j * RMD_ULONG_SIZE_T) + __BBYTE]] ) % + ( UCHAR_MAX + 1 ) ) * + data_tm[(j * RMD_ULONG_SIZE_T) + __ABYTE]) / UCHAR_MAX; + + if ((k % 2) && (i % 2)) { + int idx = x_2 + (i - x_offset) / 2 + ((k - y_offset) / 2 + y_2) * yuv->uv_width; + + avg3 = avg_4_pixels( data_tm, + (width_tm + column_discard_stride), + k, i, __ABYTE); + avg2 = avg_4_pixels( data_tm, + (width_tm + column_discard_stride), + k, i, __RBYTE); + avg1 = avg_4_pixels( data_tm, + (width_tm + column_discard_stride), + k, i, __GBYTE); + avg0 = avg_4_pixels( data_tm, + (width_tm + column_discard_stride), + k, i, __BBYTE); + + yuv->u[idx] = + (yuv->u[idx] * (UCHAR_MAX - avg3) + + ((Ur[avg2] + Ug[avg1] + UbVr[avg0]) % (UCHAR_MAX + 1)) + * avg3) / UCHAR_MAX; + + yuv->v[idx]= + (yuv->u[idx] * (UCHAR_MAX - avg3) + + ((UbVr[avg2] + Vg[avg1] + Vb[avg0]) % (UCHAR_MAX + 1)) + * avg3) / UCHAR_MAX; + } + } + } +} |