| 1 | /* |
| 2 | * Copyright (c) 2010 Niel van der Westhuizen <nielkie@gmail.com> |
| 3 | * Copyright (c) 2002 A'rpi |
| 4 | * Copyright (c) 1997-2001 ZSNES Team ( zsknight@zsnes.com / _demo_@zsnes.com ) |
| 5 | * |
| 6 | * This file is part of FFmpeg. |
| 7 | * |
| 8 | * FFmpeg is free software; you can redistribute it and/or modify |
| 9 | * it under the terms of the GNU General Public License as published by |
| 10 | * the Free Software Foundation; either version 2 of the License, or |
| 11 | * (at your option) any later version. |
| 12 | * |
| 13 | * FFmpeg is distributed in the hope that it will be useful, |
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | * GNU General Public License for more details. |
| 17 | * |
| 18 | * You should have received a copy of the GNU General Public License along |
| 19 | * with FFmpeg; if not, write to the Free Software Foundation, Inc., |
| 20 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 21 | */ |
| 22 | |
| 23 | /** |
| 24 | * @file |
| 25 | * Super 2xSaI video filter |
| 26 | * Ported from MPlayer libmpcodecs/vf_2xsai.c. |
| 27 | */ |
| 28 | |
| 29 | #include "libavutil/pixdesc.h" |
| 30 | #include "libavutil/intreadwrite.h" |
| 31 | #include "avfilter.h" |
| 32 | #include "formats.h" |
| 33 | #include "internal.h" |
| 34 | #include "video.h" |
| 35 | |
| 36 | typedef struct { |
| 37 | /* masks used for two pixels interpolation */ |
| 38 | uint32_t hi_pixel_mask; |
| 39 | uint32_t lo_pixel_mask; |
| 40 | |
| 41 | /* masks used for four pixels interpolation */ |
| 42 | uint32_t q_hi_pixel_mask; |
| 43 | uint32_t q_lo_pixel_mask; |
| 44 | |
| 45 | int bpp; ///< bytes per pixel, pixel stride for each (packed) pixel |
| 46 | int is_be; |
| 47 | } Super2xSaIContext; |
| 48 | |
| 49 | #define GET_RESULT(A, B, C, D) ((A != C || A != D) - (B != C || B != D)) |
| 50 | |
| 51 | #define INTERPOLATE(A, B) (((A & hi_pixel_mask) >> 1) + ((B & hi_pixel_mask) >> 1) + (A & B & lo_pixel_mask)) |
| 52 | |
| 53 | #define Q_INTERPOLATE(A, B, C, D) ((A & q_hi_pixel_mask) >> 2) + ((B & q_hi_pixel_mask) >> 2) + ((C & q_hi_pixel_mask) >> 2) + ((D & q_hi_pixel_mask) >> 2) \ |
| 54 | + ((((A & q_lo_pixel_mask) + (B & q_lo_pixel_mask) + (C & q_lo_pixel_mask) + (D & q_lo_pixel_mask)) >> 2) & q_lo_pixel_mask) |
| 55 | |
| 56 | static void super2xsai(AVFilterContext *ctx, |
| 57 | uint8_t *src, int src_linesize, |
| 58 | uint8_t *dst, int dst_linesize, |
| 59 | int width, int height) |
| 60 | { |
| 61 | Super2xSaIContext *sai = ctx->priv; |
| 62 | unsigned int x, y; |
| 63 | uint32_t color[4][4]; |
| 64 | unsigned char *src_line[4]; |
| 65 | const int bpp = sai->bpp; |
| 66 | const uint32_t hi_pixel_mask = sai->hi_pixel_mask; |
| 67 | const uint32_t lo_pixel_mask = sai->lo_pixel_mask; |
| 68 | const uint32_t q_hi_pixel_mask = sai->q_hi_pixel_mask; |
| 69 | const uint32_t q_lo_pixel_mask = sai->q_lo_pixel_mask; |
| 70 | |
| 71 | /* Point to the first 4 lines, first line is duplicated */ |
| 72 | src_line[0] = src; |
| 73 | src_line[1] = src; |
| 74 | src_line[2] = src + src_linesize*FFMIN(1, height-1); |
| 75 | src_line[3] = src + src_linesize*FFMIN(2, height-1); |
| 76 | |
| 77 | #define READ_COLOR4(dst, src_line, off) dst = *((const uint32_t *)src_line + off) |
| 78 | #define READ_COLOR3(dst, src_line, off) dst = AV_RL24 (src_line + 3*off) |
| 79 | #define READ_COLOR2(dst, src_line, off) dst = sai->is_be ? AV_RB16(src_line + 2 * off) : AV_RL16(src_line + 2 * off) |
| 80 | |
| 81 | for (y = 0; y < height; y++) { |
| 82 | uint8_t *dst_line[2]; |
| 83 | |
| 84 | dst_line[0] = dst + dst_linesize*2*y; |
| 85 | dst_line[1] = dst + dst_linesize*(2*y+1); |
| 86 | |
| 87 | switch (bpp) { |
| 88 | case 4: |
| 89 | READ_COLOR4(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR4(color[0][2], src_line[0], 1); READ_COLOR4(color[0][3], src_line[0], 2); |
| 90 | READ_COLOR4(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR4(color[1][2], src_line[1], 1); READ_COLOR4(color[1][3], src_line[1], 2); |
| 91 | READ_COLOR4(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR4(color[2][2], src_line[2], 1); READ_COLOR4(color[2][3], src_line[2], 2); |
| 92 | READ_COLOR4(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR4(color[3][2], src_line[3], 1); READ_COLOR4(color[3][3], src_line[3], 2); |
| 93 | break; |
| 94 | case 3: |
| 95 | READ_COLOR3(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR3(color[0][2], src_line[0], 1); READ_COLOR3(color[0][3], src_line[0], 2); |
| 96 | READ_COLOR3(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR3(color[1][2], src_line[1], 1); READ_COLOR3(color[1][3], src_line[1], 2); |
| 97 | READ_COLOR3(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR3(color[2][2], src_line[2], 1); READ_COLOR3(color[2][3], src_line[2], 2); |
| 98 | READ_COLOR3(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR3(color[3][2], src_line[3], 1); READ_COLOR3(color[3][3], src_line[3], 2); |
| 99 | break; |
| 100 | default: |
| 101 | READ_COLOR2(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR2(color[0][2], src_line[0], 1); READ_COLOR2(color[0][3], src_line[0], 2); |
| 102 | READ_COLOR2(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR2(color[1][2], src_line[1], 1); READ_COLOR2(color[1][3], src_line[1], 2); |
| 103 | READ_COLOR2(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR2(color[2][2], src_line[2], 1); READ_COLOR2(color[2][3], src_line[2], 2); |
| 104 | READ_COLOR2(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR2(color[3][2], src_line[3], 1); READ_COLOR2(color[3][3], src_line[3], 2); |
| 105 | } |
| 106 | |
| 107 | for (x = 0; x < width; x++) { |
| 108 | uint32_t product1a, product1b, product2a, product2b; |
| 109 | |
| 110 | //--------------------------------------- B0 B1 B2 B3 0 1 2 3 |
| 111 | // 4 5* 6 S2 -> 4 5* 6 7 |
| 112 | // 1 2 3 S1 8 9 10 11 |
| 113 | // A0 A1 A2 A3 12 13 14 15 |
| 114 | //-------------------------------------- |
| 115 | if (color[2][1] == color[1][2] && color[1][1] != color[2][2]) { |
| 116 | product2b = color[2][1]; |
| 117 | product1b = product2b; |
| 118 | } else if (color[1][1] == color[2][2] && color[2][1] != color[1][2]) { |
| 119 | product2b = color[1][1]; |
| 120 | product1b = product2b; |
| 121 | } else if (color[1][1] == color[2][2] && color[2][1] == color[1][2]) { |
| 122 | int r = 0; |
| 123 | |
| 124 | r += GET_RESULT(color[1][2], color[1][1], color[1][0], color[3][1]); |
| 125 | r += GET_RESULT(color[1][2], color[1][1], color[2][0], color[0][1]); |
| 126 | r += GET_RESULT(color[1][2], color[1][1], color[3][2], color[2][3]); |
| 127 | r += GET_RESULT(color[1][2], color[1][1], color[0][2], color[1][3]); |
| 128 | |
| 129 | if (r > 0) |
| 130 | product1b = color[1][2]; |
| 131 | else if (r < 0) |
| 132 | product1b = color[1][1]; |
| 133 | else |
| 134 | product1b = INTERPOLATE(color[1][1], color[1][2]); |
| 135 | |
| 136 | product2b = product1b; |
| 137 | } else { |
| 138 | if (color[1][2] == color[2][2] && color[2][2] == color[3][1] && color[2][1] != color[3][2] && color[2][2] != color[3][0]) |
| 139 | product2b = Q_INTERPOLATE(color[2][2], color[2][2], color[2][2], color[2][1]); |
| 140 | else if (color[1][1] == color[2][1] && color[2][1] == color[3][2] && color[3][1] != color[2][2] && color[2][1] != color[3][3]) |
| 141 | product2b = Q_INTERPOLATE(color[2][1], color[2][1], color[2][1], color[2][2]); |
| 142 | else |
| 143 | product2b = INTERPOLATE(color[2][1], color[2][2]); |
| 144 | |
| 145 | if (color[1][2] == color[2][2] && color[1][2] == color[0][1] && color[1][1] != color[0][2] && color[1][2] != color[0][0]) |
| 146 | product1b = Q_INTERPOLATE(color[1][2], color[1][2], color[1][2], color[1][1]); |
| 147 | else if (color[1][1] == color[2][1] && color[1][1] == color[0][2] && color[0][1] != color[1][2] && color[1][1] != color[0][3]) |
| 148 | product1b = Q_INTERPOLATE(color[1][2], color[1][1], color[1][1], color[1][1]); |
| 149 | else |
| 150 | product1b = INTERPOLATE(color[1][1], color[1][2]); |
| 151 | } |
| 152 | |
| 153 | if (color[1][1] == color[2][2] && color[2][1] != color[1][2] && color[1][0] == color[1][1] && color[1][1] != color[3][2]) |
| 154 | product2a = INTERPOLATE(color[2][1], color[1][1]); |
| 155 | else if (color[1][1] == color[2][0] && color[1][2] == color[1][1] && color[1][0] != color[2][1] && color[1][1] != color[3][0]) |
| 156 | product2a = INTERPOLATE(color[2][1], color[1][1]); |
| 157 | else |
| 158 | product2a = color[2][1]; |
| 159 | |
| 160 | if (color[2][1] == color[1][2] && color[1][1] != color[2][2] && color[2][0] == color[2][1] && color[2][1] != color[0][2]) |
| 161 | product1a = INTERPOLATE(color[2][1], color[1][1]); |
| 162 | else if (color[1][0] == color[2][1] && color[2][2] == color[2][1] && color[2][0] != color[1][1] && color[2][1] != color[0][0]) |
| 163 | product1a = INTERPOLATE(color[2][1], color[1][1]); |
| 164 | else |
| 165 | product1a = color[1][1]; |
| 166 | |
| 167 | /* Set the calculated pixels */ |
| 168 | switch (bpp) { |
| 169 | case 4: |
| 170 | AV_WN32A(dst_line[0] + x * 8, product1a); |
| 171 | AV_WN32A(dst_line[0] + x * 8 + 4, product1b); |
| 172 | AV_WN32A(dst_line[1] + x * 8, product2a); |
| 173 | AV_WN32A(dst_line[1] + x * 8 + 4, product2b); |
| 174 | break; |
| 175 | case 3: |
| 176 | AV_WL24(dst_line[0] + x * 6, product1a); |
| 177 | AV_WL24(dst_line[0] + x * 6 + 3, product1b); |
| 178 | AV_WL24(dst_line[1] + x * 6, product2a); |
| 179 | AV_WL24(dst_line[1] + x * 6 + 3, product2b); |
| 180 | break; |
| 181 | default: // bpp = 2 |
| 182 | if (sai->is_be) { |
| 183 | AV_WB32(dst_line[0] + x * 4, product1a | (product1b << 16)); |
| 184 | AV_WB32(dst_line[1] + x * 4, product2a | (product2b << 16)); |
| 185 | } else { |
| 186 | AV_WL32(dst_line[0] + x * 4, product1a | (product1b << 16)); |
| 187 | AV_WL32(dst_line[1] + x * 4, product2a | (product2b << 16)); |
| 188 | } |
| 189 | } |
| 190 | |
| 191 | /* Move color matrix forward */ |
| 192 | color[0][0] = color[0][1]; color[0][1] = color[0][2]; color[0][2] = color[0][3]; |
| 193 | color[1][0] = color[1][1]; color[1][1] = color[1][2]; color[1][2] = color[1][3]; |
| 194 | color[2][0] = color[2][1]; color[2][1] = color[2][2]; color[2][2] = color[2][3]; |
| 195 | color[3][0] = color[3][1]; color[3][1] = color[3][2]; color[3][2] = color[3][3]; |
| 196 | |
| 197 | if (x < width - 3) { |
| 198 | x += 3; |
| 199 | switch (bpp) { |
| 200 | case 4: |
| 201 | READ_COLOR4(color[0][3], src_line[0], x); |
| 202 | READ_COLOR4(color[1][3], src_line[1], x); |
| 203 | READ_COLOR4(color[2][3], src_line[2], x); |
| 204 | READ_COLOR4(color[3][3], src_line[3], x); |
| 205 | break; |
| 206 | case 3: |
| 207 | READ_COLOR3(color[0][3], src_line[0], x); |
| 208 | READ_COLOR3(color[1][3], src_line[1], x); |
| 209 | READ_COLOR3(color[2][3], src_line[2], x); |
| 210 | READ_COLOR3(color[3][3], src_line[3], x); |
| 211 | break; |
| 212 | default: /* case 2 */ |
| 213 | READ_COLOR2(color[0][3], src_line[0], x); |
| 214 | READ_COLOR2(color[1][3], src_line[1], x); |
| 215 | READ_COLOR2(color[2][3], src_line[2], x); |
| 216 | READ_COLOR2(color[3][3], src_line[3], x); |
| 217 | } |
| 218 | x -= 3; |
| 219 | } |
| 220 | } |
| 221 | |
| 222 | /* We're done with one line, so we shift the source lines up */ |
| 223 | src_line[0] = src_line[1]; |
| 224 | src_line[1] = src_line[2]; |
| 225 | src_line[2] = src_line[3]; |
| 226 | |
| 227 | /* Read next line */ |
| 228 | src_line[3] = src_line[2]; |
| 229 | if (y < height - 3) |
| 230 | src_line[3] += src_linesize; |
| 231 | } // y loop |
| 232 | } |
| 233 | |
| 234 | static int query_formats(AVFilterContext *ctx) |
| 235 | { |
| 236 | static const enum AVPixelFormat pix_fmts[] = { |
| 237 | AV_PIX_FMT_RGBA, AV_PIX_FMT_BGRA, AV_PIX_FMT_ARGB, AV_PIX_FMT_ABGR, |
| 238 | AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24, |
| 239 | AV_PIX_FMT_RGB565BE, AV_PIX_FMT_BGR565BE, AV_PIX_FMT_RGB555BE, AV_PIX_FMT_BGR555BE, |
| 240 | AV_PIX_FMT_RGB565LE, AV_PIX_FMT_BGR565LE, AV_PIX_FMT_RGB555LE, AV_PIX_FMT_BGR555LE, |
| 241 | AV_PIX_FMT_NONE |
| 242 | }; |
| 243 | |
| 244 | ff_set_common_formats(ctx, ff_make_format_list(pix_fmts)); |
| 245 | return 0; |
| 246 | } |
| 247 | |
| 248 | static int config_input(AVFilterLink *inlink) |
| 249 | { |
| 250 | Super2xSaIContext *sai = inlink->dst->priv; |
| 251 | |
| 252 | sai->hi_pixel_mask = 0xFEFEFEFE; |
| 253 | sai->lo_pixel_mask = 0x01010101; |
| 254 | sai->q_hi_pixel_mask = 0xFCFCFCFC; |
| 255 | sai->q_lo_pixel_mask = 0x03030303; |
| 256 | sai->bpp = 4; |
| 257 | |
| 258 | switch (inlink->format) { |
| 259 | case AV_PIX_FMT_RGB24: |
| 260 | case AV_PIX_FMT_BGR24: |
| 261 | sai->bpp = 3; |
| 262 | break; |
| 263 | |
| 264 | case AV_PIX_FMT_RGB565BE: |
| 265 | case AV_PIX_FMT_BGR565BE: |
| 266 | sai->is_be = 1; |
| 267 | case AV_PIX_FMT_RGB565LE: |
| 268 | case AV_PIX_FMT_BGR565LE: |
| 269 | sai->hi_pixel_mask = 0xF7DEF7DE; |
| 270 | sai->lo_pixel_mask = 0x08210821; |
| 271 | sai->q_hi_pixel_mask = 0xE79CE79C; |
| 272 | sai->q_lo_pixel_mask = 0x18631863; |
| 273 | sai->bpp = 2; |
| 274 | break; |
| 275 | |
| 276 | case AV_PIX_FMT_BGR555BE: |
| 277 | case AV_PIX_FMT_RGB555BE: |
| 278 | sai->is_be = 1; |
| 279 | case AV_PIX_FMT_BGR555LE: |
| 280 | case AV_PIX_FMT_RGB555LE: |
| 281 | sai->hi_pixel_mask = 0x7BDE7BDE; |
| 282 | sai->lo_pixel_mask = 0x04210421; |
| 283 | sai->q_hi_pixel_mask = 0x739C739C; |
| 284 | sai->q_lo_pixel_mask = 0x0C630C63; |
| 285 | sai->bpp = 2; |
| 286 | break; |
| 287 | } |
| 288 | |
| 289 | return 0; |
| 290 | } |
| 291 | |
| 292 | static int config_output(AVFilterLink *outlink) |
| 293 | { |
| 294 | AVFilterLink *inlink = outlink->src->inputs[0]; |
| 295 | |
| 296 | outlink->w = inlink->w*2; |
| 297 | outlink->h = inlink->h*2; |
| 298 | |
| 299 | av_log(inlink->dst, AV_LOG_VERBOSE, "fmt:%s size:%dx%d -> size:%dx%d\n", |
| 300 | av_get_pix_fmt_name(inlink->format), |
| 301 | inlink->w, inlink->h, outlink->w, outlink->h); |
| 302 | |
| 303 | return 0; |
| 304 | } |
| 305 | |
| 306 | static int filter_frame(AVFilterLink *inlink, AVFrame *inpicref) |
| 307 | { |
| 308 | AVFilterLink *outlink = inlink->dst->outputs[0]; |
| 309 | AVFrame *outpicref = ff_get_video_buffer(outlink, outlink->w, outlink->h); |
| 310 | if (!outpicref) { |
| 311 | av_frame_free(&inpicref); |
| 312 | return AVERROR(ENOMEM); |
| 313 | } |
| 314 | av_frame_copy_props(outpicref, inpicref); |
| 315 | outpicref->width = outlink->w; |
| 316 | outpicref->height = outlink->h; |
| 317 | |
| 318 | super2xsai(inlink->dst, inpicref->data[0], inpicref->linesize[0], |
| 319 | outpicref->data[0], outpicref->linesize[0], |
| 320 | inlink->w, inlink->h); |
| 321 | |
| 322 | av_frame_free(&inpicref); |
| 323 | return ff_filter_frame(outlink, outpicref); |
| 324 | } |
| 325 | |
| 326 | static const AVFilterPad super2xsai_inputs[] = { |
| 327 | { |
| 328 | .name = "default", |
| 329 | .type = AVMEDIA_TYPE_VIDEO, |
| 330 | .config_props = config_input, |
| 331 | .filter_frame = filter_frame, |
| 332 | }, |
| 333 | { NULL } |
| 334 | }; |
| 335 | |
| 336 | static const AVFilterPad super2xsai_outputs[] = { |
| 337 | { |
| 338 | .name = "default", |
| 339 | .type = AVMEDIA_TYPE_VIDEO, |
| 340 | .config_props = config_output, |
| 341 | }, |
| 342 | { NULL } |
| 343 | }; |
| 344 | |
| 345 | AVFilter ff_vf_super2xsai = { |
| 346 | .name = "super2xsai", |
| 347 | .description = NULL_IF_CONFIG_SMALL("Scale the input by 2x using the Super2xSaI pixel art algorithm."), |
| 348 | .priv_size = sizeof(Super2xSaIContext), |
| 349 | .query_formats = query_formats, |
| 350 | .inputs = super2xsai_inputs, |
| 351 | .outputs = super2xsai_outputs, |
| 352 | }; |