git.sesse.net Git - movit/blob - ycbcr_input.cpp

   1 #include <Eigen/Core>
   2 #include <Eigen/LU>
   3 #include <epoxy/gl.h>
   4 #include <assert.h>
   5 #include <stdio.h>
   6 #include <string.h>
   7
   8 #include "effect_util.h"
   9 #include "resource_pool.h"
  10 #include "util.h"
  11 #include "ycbcr_input.h"
  12
  13 using namespace Eigen;
  14 using namespace std;
  15
  16 namespace movit {
  17
  18 namespace {
  19
  20 // OpenGL has texel center in (0.5, 0.5), but different formats have
  21 // chroma in various other places. If luma samples are X, the chroma
  22 // sample is *, and subsampling is 3x3, the situation with chroma
  23 // center in (0.5, 0.5) looks approximately like this:
  24 //
  25 //   X   X
  26 //     *
  27 //   X   X
  28 //
  29 // If, on the other hand, chroma center is in (0.0, 0.5) (common
  30 // for e.g. MPEG-4), the figure changes to:
  31 //
  32 //   X   X
  33 //   *
  34 //   X   X
  35 //
  36 // In other words, (0.0, 0.0) means that the chroma sample is exactly
  37 // co-sited on top of the top-left luma sample. Note, however, that
  38 // this is _not_ 0.5 texels to the left, since the OpenGL's texel center
  39 // is in (0.5, 0.5); it is in (0.25, 0.25). In a sense, the four luma samples
  40 // define a square where chroma position (0.0, 0.0) is in texel position
  41 // (0.25, 0.25) and chroma position (1.0, 1.0) is in texel position (0.75, 0.75)
  42 // (the outer border shows the borders of the texel itself, ie. from
  43 // (0, 0) to (1, 1)):
  44 //
  45 //  ---------
  46 // |         |
  47 // |  X---X  |
  48 // |  | * |  |
  49 // |  X---X  |
  50 // |         |
  51 //  ---------
  52 //
  53 // Also note that if we have no subsampling, the square will have zero
  54 // area and the chroma position does not matter at all.
  55 float compute_chroma_offset(float pos, unsigned subsampling_factor, unsigned resolution)
  56 {
  57         float local_chroma_pos = (0.5 + pos * (subsampling_factor - 1)) / subsampling_factor;
  58         return (0.5 - local_chroma_pos) / resolution;
  59 }
  60
  61 // Given <ycbcr_format>, compute the values needed to turn Y'CbCr into R'G'B';
  62 // first subtract the returned offset, then left-multiply the returned matrix
  63 // (the scaling is already folded into it).
  64 void compute_ycbcr_matrix(YCbCrFormat ycbcr_format, float* offset, Matrix3d* ycbcr_to_rgb)
  65 {
  66         double coeff[3], scale[3];
  67
  68         switch (ycbcr_format.luma_coefficients) {
  69         case YCBCR_REC_601:
  70                 // Rec. 601, page 2.
  71                 coeff[0] = 0.299;
  72                 coeff[1] = 0.587;
  73                 coeff[2] = 0.114;
  74                 break;
  75
  76         case YCBCR_REC_709:
  77                 // Rec. 709, page 19.
  78                 coeff[0] = 0.2126;
  79                 coeff[1] = 0.7152;
  80                 coeff[2] = 0.0722;
  81                 break;
  82
  83         case YCBCR_REC_2020:
  84                 // Rec. 2020, page 4.
  85                 coeff[0] = 0.2627;
  86                 coeff[1] = 0.6780;
  87                 coeff[2] = 0.0593;
  88                 break;
  89
  90         default:
  91                 assert(false);
  92         }
  93
  94         if (ycbcr_format.full_range) {
  95                 offset[0] = 0.0 / 255.0;
  96                 offset[1] = 128.0 / 255.0;
  97                 offset[2] = 128.0 / 255.0;
  98
  99                 scale[0] = 1.0;
 100                 scale[1] = 1.0;
 101                 scale[2] = 1.0;
 102         } else {
 103                 // Rec. 601, page 4; Rec. 709, page 19; Rec. 2020, page 4.
 104                 offset[0] = 16.0 / 255.0;
 105                 offset[1] = 128.0 / 255.0;
 106                 offset[2] = 128.0 / 255.0;
 107
 108                 scale[0] = 255.0 / 219.0;
 109                 scale[1] = 255.0 / 224.0;
 110                 scale[2] = 255.0 / 224.0;
 111         }
 112
 113         // Matrix to convert RGB to YCbCr. See e.g. Rec. 601.
 114         Matrix3d rgb_to_ycbcr;
 115         rgb_to_ycbcr(0,0) = coeff[0];
 116         rgb_to_ycbcr(0,1) = coeff[1];
 117         rgb_to_ycbcr(0,2) = coeff[2];
 118
 119         float cb_fac = (224.0 / 219.0) / (coeff[0] + coeff[1] + 1.0f - coeff[2]);
 120         rgb_to_ycbcr(1,0) = -coeff[0] * cb_fac;
 121         rgb_to_ycbcr(1,1) = -coeff[1] * cb_fac;
 122         rgb_to_ycbcr(1,2) = (1.0f - coeff[2]) * cb_fac;
 123
 124         float cr_fac = (224.0 / 219.0) / (1.0f - coeff[0] + coeff[1] + coeff[2]);
 125         rgb_to_ycbcr(2,0) = (1.0f - coeff[0]) * cr_fac;
 126         rgb_to_ycbcr(2,1) = -coeff[1] * cr_fac;
 127         rgb_to_ycbcr(2,2) = -coeff[2] * cr_fac;
 128
 129         // Inverting the matrix gives us what we need to go from YCbCr back to RGB.
 130         *ycbcr_to_rgb = rgb_to_ycbcr.inverse();
 131
 132         // Fold in the scaling.
 133         *ycbcr_to_rgb *= Map<const Vector3d>(scale).asDiagonal();
 134 }
 135
 136 }  // namespace
 137
 138 YCbCrInput::YCbCrInput(const ImageFormat &image_format,
 139                        const YCbCrFormat &ycbcr_format,
 140                        unsigned width, unsigned height)
 141         : image_format(image_format),
 142           ycbcr_format(ycbcr_format),
 143           width(width),
 144           height(height),
 145           resource_pool(NULL)
 146 {
 147         pbos[0] = pbos[1] = pbos[2] = 0;
 148         texture_num[0] = texture_num[1] = texture_num[2] = 0;
 149
 150         assert(width % ycbcr_format.chroma_subsampling_x == 0);
 151         pitch[0] = widths[0] = width;
 152         pitch[1] = widths[1] = width / ycbcr_format.chroma_subsampling_x;
 153         pitch[2] = widths[2] = width / ycbcr_format.chroma_subsampling_x;
 154
 155         assert(height % ycbcr_format.chroma_subsampling_y == 0);
 156         heights[0] = height;
 157         heights[1] = height / ycbcr_format.chroma_subsampling_y;
 158         heights[2] = height / ycbcr_format.chroma_subsampling_y;
 159
 160         pixel_data[0] = pixel_data[1] = pixel_data[2] = NULL;
 161 }
 162
 163 YCbCrInput::~YCbCrInput()
 164 {
 165         for (unsigned channel = 0; channel < 3; ++channel) {
 166                 if (texture_num[channel] != 0) {
 167                         resource_pool->release_2d_texture(texture_num[channel]);
 168                 }
 169         }
 170 }
 171
 172 void YCbCrInput::set_gl_state(GLuint glsl_program_num, const string& prefix, unsigned *sampler_num)
 173 {
 174         for (unsigned channel = 0; channel < 3; ++channel) {
 175                 glActiveTexture(GL_TEXTURE0 + *sampler_num + channel);
 176                 check_error();
 177
 178                 if (texture_num[channel] == 0) {
 179                         // (Re-)upload the texture.
 180                         texture_num[channel] = resource_pool->create_2d_texture(GL_R8, widths[channel], heights[channel]);
 181                         glBindTexture(GL_TEXTURE_2D, texture_num[channel]);
 182                         check_error();
 183                         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
 184                         check_error();
 185                         glBindBuffer(GL_PIXEL_UNPACK_BUFFER_ARB, pbos[channel]);
 186                         check_error();
 187                         glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
 188                         check_error();
 189                         glPixelStorei(GL_UNPACK_ROW_LENGTH, pitch[channel]);
 190                         check_error();
 191                         glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, widths[channel], heights[channel], GL_RED, GL_UNSIGNED_BYTE, pixel_data[channel]);
 192                         check_error();
 193                         glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
 194                         check_error();
 195                         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
 196                         check_error();
 197                         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 198                         check_error();
 199                 } else {
 200                         glBindTexture(GL_TEXTURE_2D, texture_num[channel]);
 201                         check_error();
 202                 }
 203         }
 204
 205         glBindBuffer(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
 206         check_error();
 207
 208         // Bind samplers.
 209         set_uniform_int(glsl_program_num, prefix, "tex_y", *sampler_num + 0);
 210         set_uniform_int(glsl_program_num, prefix, "tex_cb", *sampler_num + 1);
 211         set_uniform_int(glsl_program_num, prefix, "tex_cr", *sampler_num + 2);
 212
 213         *sampler_num += 3;
 214 }
 215
 216 string YCbCrInput::output_fragment_shader()
 217 {
 218         float offset[3];
 219         Matrix3d ycbcr_to_rgb;
 220         compute_ycbcr_matrix(ycbcr_format, offset, &ycbcr_to_rgb);
 221
 222         string frag_shader;
 223
 224         frag_shader = output_glsl_mat3("PREFIX(inv_ycbcr_matrix)", ycbcr_to_rgb);
 225         frag_shader += output_glsl_vec3("PREFIX(offset)", offset[0], offset[1], offset[2]);
 226
 227         float cb_offset_x = compute_chroma_offset(
 228                 ycbcr_format.cb_x_position, ycbcr_format.chroma_subsampling_x, widths[1]);
 229         float cb_offset_y = compute_chroma_offset(
 230                 ycbcr_format.cb_y_position, ycbcr_format.chroma_subsampling_y, heights[1]);
 231         frag_shader += output_glsl_vec2("PREFIX(cb_offset)", cb_offset_x, cb_offset_y);
 232
 233         float cr_offset_x = compute_chroma_offset(
 234                 ycbcr_format.cr_x_position, ycbcr_format.chroma_subsampling_x, widths[2]);
 235         float cr_offset_y = compute_chroma_offset(
 236                 ycbcr_format.cr_y_position, ycbcr_format.chroma_subsampling_y, heights[2]);
 237         frag_shader += output_glsl_vec2("PREFIX(cr_offset)", cr_offset_x, cr_offset_y);
 238
 239         frag_shader += read_file("ycbcr_input.frag");
 240         return frag_shader;
 241 }
 242
 243 void YCbCrInput::invalidate_pixel_data()
 244 {
 245         for (unsigned channel = 0; channel < 3; ++channel) {
 246                 if (texture_num[channel] != 0) {
 247                         resource_pool->release_2d_texture(texture_num[channel]);
 248                         texture_num[channel] = 0;
 249                 }
 250         }
 251 }
 252
 253 bool YCbCrInput::set_int(const std::string& key, int value)
 254 {
 255         if (key == "needs_mipmaps") {
 256                 // We currently do not support this.
 257                 return (value == 0);
 258         }
 259         return Effect::set_int(key, value);
 260 }
 261
 262 }  // namespace movit