git.sesse.net Git - ffmpeg/blob - libavfilter/vf_lenscorrection.c

   1 /*
   2  * Copyright (C) 2007 Richard Spindler (author of frei0r plugin from which this was derived)
   3  * Copyright (C) 2014 Daniel Oberhoff
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * Lenscorrection filter, algorithm from the frei0r plugin with the same name
  25 */
  26 #include <stdlib.h>
  27 #include <math.h>
  28
  29 #include "libavutil/colorspace.h"
  30 #include "libavutil/opt.h"
  31 #include "libavutil/intreadwrite.h"
  32 #include "libavutil/pixdesc.h"
  33
  34 #include "avfilter.h"
  35 #include "drawutils.h"
  36 #include "internal.h"
  37 #include "video.h"
  38
  39 typedef struct LenscorrectionCtx {
  40     const AVClass *av_class;
  41     int width;
  42     int height;
  43     int hsub, vsub;
  44     int depth;
  45     int nb_planes;
  46     double cx, cy, k1, k2;
  47     int interpolation;
  48     uint8_t fill_rgba[4];
  49     int fill_color[4];
  50
  51     int32_t *correction[4];
  52
  53     int (*filter_slice)(AVFilterContext *ctx, void *arg, int job, int nb_jobs);
  54 } LenscorrectionCtx;
  55
  56 #define OFFSET(x) offsetof(LenscorrectionCtx, x)
  57 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
  58 static const AVOption lenscorrection_options[] = {
  59     { "cx", "set relative center x", OFFSET(cx), AV_OPT_TYPE_DOUBLE, {.dbl=0.5}, 0, 1, .flags=FLAGS },
  60     { "cy", "set relative center y", OFFSET(cy), AV_OPT_TYPE_DOUBLE, {.dbl=0.5}, 0, 1, .flags=FLAGS },
  61     { "k1", "set quadratic distortion factor", OFFSET(k1), AV_OPT_TYPE_DOUBLE, {.dbl=0.0}, -1, 1, .flags=FLAGS },
  62     { "k2", "set double quadratic distortion factor", OFFSET(k2), AV_OPT_TYPE_DOUBLE, {.dbl=0.0}, -1, 1, .flags=FLAGS },
  63     { "i",  "set interpolation type", OFFSET(interpolation), AV_OPT_TYPE_INT, {.i64=0}, 0, 64, .flags=FLAGS, "i" },
  64     {  "nearest",  "nearest neighbour", 0,                   AV_OPT_TYPE_CONST, {.i64=0},0, 0, .flags=FLAGS, "i" },
  65     {  "bilinear", "bilinear",          0,                   AV_OPT_TYPE_CONST, {.i64=1},0, 0, .flags=FLAGS, "i" },
  66     { "fc", "set the color of the unmapped pixels", OFFSET(fill_rgba), AV_OPT_TYPE_COLOR, {.str="black@0"}, .flags = FLAGS },
  67     { NULL }
  68 };
  69
  70 AVFILTER_DEFINE_CLASS(lenscorrection);
  71
  72 typedef struct ThreadData {
  73     AVFrame *in, *out;
  74     int w, h;
  75     int depth;
  76     int plane;
  77     int fill_color;
  78     int xcenter, ycenter;
  79     int32_t *correction;
  80 } ThreadData;
  81
  82 #define NEAREST(type, name)                                                    \
  83 static int filter##name##_slice(AVFilterContext *ctx, void *arg, int job,      \
  84                                 int nb_jobs)                                   \
  85 {                                                                              \
  86     ThreadData *td = arg;                                                      \
  87     AVFrame *in = td->in;                                                      \
  88     AVFrame *out = td->out;                                                    \
  89                                                                                \
  90     const int fill_color = td->fill_color;                                     \
  91     const int w = td->w, h = td->h;                                            \
  92     const int xcenter = td->xcenter;                                           \
  93     const int ycenter = td->ycenter;                                           \
  94     const int start = (h *  job   ) / nb_jobs;                                 \
  95     const int end   = (h * (job+1)) / nb_jobs;                                 \
  96     const int plane = td->plane;                                               \
  97     const int inlinesize = in->linesize[plane] / sizeof(type);                 \
  98     const int outlinesize = out->linesize[plane] / sizeof(type);               \
  99     const type *indata = (const type *)in->data[plane];                        \
 100     type *outrow = (type *)out->data[plane] + start * outlinesize;             \
 101     for (int i = start; i < end; i++, outrow += outlinesize) {                 \
 102         const int off_y = i - ycenter;                                         \
 103         type *out = outrow;                                                    \
 104         for (int j = 0; j < w; j++) {                                          \
 105             const int off_x = j - xcenter;                                     \
 106             const int64_t radius_mult = td->correction[j + i*w];               \
 107             const int x = xcenter + ((radius_mult * off_x + (1<<23))>>24);     \
 108             const int y = ycenter + ((radius_mult * off_y + (1<<23))>>24);     \
 109             const char isvalid = x >= 0 && x < w && y >= 0 && y < h;           \
 110             *out++ =  isvalid ? indata[y * inlinesize + x] : fill_color;       \
 111         }                                                                      \
 112     }                                                                          \
 113     return 0;                                                                  \
 114 }
 115
 116
 117 NEAREST(uint8_t, 8)
 118 NEAREST(uint16_t, 16)
 119
 120 #define BILINEAR(type, name)                                                   \
 121 static int filter##name##_slice_bilinear(AVFilterContext *ctx, void *arg,      \
 122                                          int job, int nb_jobs)                 \
 123 {                                                                              \
 124     ThreadData *td = arg;                                                      \
 125     AVFrame *in = td->in;                                                      \
 126     AVFrame *out = td->out;                                                    \
 127                                                                                \
 128     const int fill_color = td->fill_color;                                     \
 129     const int depth = td->depth;                                               \
 130     const uint64_t max = (1 << 24) - 1;                                        \
 131     const uint64_t add = (1 << 23);                                            \
 132     const int w = td->w, h = td->h;                                            \
 133     const int xcenter = td->xcenter;                                           \
 134     const int ycenter = td->ycenter;                                           \
 135     const int start = (h *  job   ) / nb_jobs;                                 \
 136     const int end   = (h * (job+1)) / nb_jobs;                                 \
 137     const int plane = td->plane;                                               \
 138     const int inlinesize = in->linesize[plane] / sizeof(type);                 \
 139     const int outlinesize = out->linesize[plane] / sizeof(type);               \
 140     const type *indata = (const type *)in->data[plane];                        \
 141     type *outrow = (type *)out->data[plane] + start * outlinesize;             \
 142                                                                                \
 143     for (int i = start; i < end; i++, outrow += outlinesize) {                 \
 144         const int off_y = i - ycenter;                                         \
 145         type *out = outrow;                                                    \
 146                                                                                \
 147         for (int j = 0; j < w; j++) {                                          \
 148             const int off_x = j - xcenter;                                     \
 149             const int64_t radius_mult = td->correction[j + i*w];               \
 150             const int x = xcenter + ((radius_mult * off_x + (1<<23)) >> 24);   \
 151             const int y = ycenter + ((radius_mult * off_y + (1<<23)) >> 24);   \
 152             const char isvalid = x >= 0 && x <= w - 1 && y >= 0 && y <= h - 1; \
 153                                                                                \
 154             if (isvalid) {                                                     \
 155                 const int nx = FFMIN(x + 1, w - 1);                            \
 156                 const int ny = FFMIN(y + 1, h - 1);                            \
 157                 const uint64_t du = off_x >= 0 ? (radius_mult * off_x + add) & max : max - ((radius_mult * -off_x + add) & max); \
 158                 const uint64_t dv = off_y >= 0 ? (radius_mult * off_y + add) & max : max - ((radius_mult * -off_y + add) & max); \
 159                 const uint64_t p0 = indata[ y * inlinesize +  x];              \
 160                 const uint64_t p1 = indata[ y * inlinesize + nx];              \
 161                 const uint64_t p2 = indata[ny * inlinesize +  x];              \
 162                 const uint64_t p3 = indata[ny * inlinesize + nx];              \
 163                 uint64_t sum = 0;                                              \
 164                                                                                \
 165                 sum += (max - du) * (max - dv) * p0;                           \
 166                 sum += (      du) * (max - dv) * p1;                           \
 167                 sum += (max - du) * (      dv) * p2;                           \
 168                 sum += (      du) * (      dv) * p3;                           \
 169                                                                                \
 170                 out[j] = av_clip_uintp2_c((sum + (1ULL << 47)) >> 48, depth);  \
 171             } else {                                                           \
 172                 out[j] = fill_color;                                           \
 173             }                                                                  \
 174         }                                                                      \
 175     }                                                                          \
 176                                                                                \
 177     return 0;                                                                  \
 178 }
 179
 180 BILINEAR(uint8_t, 8)
 181 BILINEAR(uint16_t, 16)
 182
 183 static int query_formats(AVFilterContext *ctx)
 184 {
 185     static const enum AVPixelFormat pix_fmts[] = {
 186         AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9,
 187         AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14,
 188         AV_PIX_FMT_GRAY16,
 189         AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P,
 190         AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
 191         AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P,
 192         AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P,
 193         AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_YUVJ444P,
 194         AV_PIX_FMT_YUVJ411P,
 195         AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
 196         AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
 197         AV_PIX_FMT_YUV440P10,
 198         AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV420P12,
 199         AV_PIX_FMT_YUV440P12,
 200         AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV420P14,
 201         AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
 202         AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10,
 203         AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
 204         AV_PIX_FMT_YUVA420P,  AV_PIX_FMT_YUVA422P,   AV_PIX_FMT_YUVA444P,
 205         AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P12, AV_PIX_FMT_YUVA444P16,
 206         AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA422P16,
 207         AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
 208         AV_PIX_FMT_GBRAP,     AV_PIX_FMT_GBRAP10,    AV_PIX_FMT_GBRAP12,    AV_PIX_FMT_GBRAP16,
 209         AV_PIX_FMT_NONE
 210     };
 211     AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
 212     if (!fmts_list)
 213         return AVERROR(ENOMEM);
 214     return ff_set_common_formats(ctx, fmts_list);
 215 }
 216
 217 static av_cold void uninit(AVFilterContext *ctx)
 218 {
 219     LenscorrectionCtx *rect = ctx->priv;
 220     int i;
 221
 222     for (i = 0; i < FF_ARRAY_ELEMS(rect->correction); i++) {
 223         av_freep(&rect->correction[i]);
 224     }
 225 }
 226
 227 static void calc_correction(AVFilterContext *ctx, int plane)
 228 {
 229     LenscorrectionCtx *rect = ctx->priv;
 230     int hsub = plane == 1 || plane == 2 ? rect->hsub : 0;
 231     int vsub = plane == 1 || plane == 2 ? rect->vsub : 0;
 232     int w = AV_CEIL_RSHIFT(rect->width, hsub);
 233     int h = AV_CEIL_RSHIFT(rect->height, vsub);
 234     int xcenter = rect->cx * w;
 235     int ycenter = rect->cy * h;
 236     int k1 = rect->k1 * (1<<24);
 237     int k2 = rect->k2 * (1<<24);
 238     const int64_t r2inv = (4LL<<60) / (w * w + h * h);
 239
 240     for (int j = 0; j < h; j++) {
 241         const int off_y = j - ycenter;
 242         const int off_y2 = off_y * off_y;
 243         for (int i = 0; i < w; i++) {
 244             const int off_x = i - xcenter;
 245             const int64_t r2 = ((off_x * off_x + off_y2) * r2inv + (1LL<<31)) >> 32;
 246             const int64_t r4 = (r2 * r2 + (1<<27)) >> 28;
 247             const int radius_mult = (r2 * k1 + r4 * k2 + (1LL<<27) + (1LL<<52))>>28;
 248             rect->correction[plane][j * w + i] = radius_mult;
 249         }
 250     }
 251 }
 252
 253 static int config_output(AVFilterLink *outlink)
 254 {
 255     AVFilterContext *ctx = outlink->src;
 256     LenscorrectionCtx *rect = ctx->priv;
 257     AVFilterLink *inlink = ctx->inputs[0];
 258     const AVPixFmtDescriptor *pixdesc = av_pix_fmt_desc_get(inlink->format);
 259     int is_rgb = !!(pixdesc->flags & AV_PIX_FMT_FLAG_RGB);
 260     uint8_t rgba_map[4];
 261     int factor;
 262
 263     ff_fill_rgba_map(rgba_map, inlink->format);
 264     rect->depth = pixdesc->comp[0].depth;
 265     factor = 1 << (rect->depth - 8);
 266     rect->hsub = pixdesc->log2_chroma_w;
 267     rect->vsub = pixdesc->log2_chroma_h;
 268     outlink->w = rect->width = inlink->w;
 269     outlink->h = rect->height = inlink->h;
 270     rect->nb_planes = av_pix_fmt_count_planes(inlink->format);
 271     rect->filter_slice = rect->depth <= 8 ? filter8_slice : filter16_slice;
 272     if (rect->interpolation)
 273         rect->filter_slice = rect->depth <= 8 ? filter8_slice_bilinear : filter16_slice_bilinear;
 274
 275     if (is_rgb) {
 276         rect->fill_color[rgba_map[0]] = rect->fill_rgba[0] * factor;
 277         rect->fill_color[rgba_map[1]] = rect->fill_rgba[1] * factor;
 278         rect->fill_color[rgba_map[2]] = rect->fill_rgba[2] * factor;
 279         rect->fill_color[rgba_map[3]] = rect->fill_rgba[3] * factor;
 280     } else {
 281         rect->fill_color[0] = RGB_TO_Y_BT709(rect->fill_rgba[0], rect->fill_rgba[1], rect->fill_rgba[2]) * factor;
 282         rect->fill_color[1] = RGB_TO_U_BT709(rect->fill_rgba[0], rect->fill_rgba[1], rect->fill_rgba[2], 0) * factor;
 283         rect->fill_color[2] = RGB_TO_V_BT709(rect->fill_rgba[0], rect->fill_rgba[1], rect->fill_rgba[2], 0) * factor;
 284         rect->fill_color[3] = rect->fill_rgba[3] * factor;
 285     }
 286
 287     for (int plane = 0; plane < rect->nb_planes; plane++) {
 288         int hsub = plane == 1 || plane == 2 ? rect->hsub : 0;
 289         int vsub = plane == 1 || plane == 2 ? rect->vsub : 0;
 290         int w = AV_CEIL_RSHIFT(rect->width, hsub);
 291         int h = AV_CEIL_RSHIFT(rect->height, vsub);
 292
 293         if (!rect->correction[plane])
 294             rect->correction[plane] = av_malloc_array(w, h * sizeof(**rect->correction));
 295         if (!rect->correction[plane])
 296             return AVERROR(ENOMEM);
 297         calc_correction(ctx, plane);
 298     }
 299
 300     return 0;
 301 }
 302
 303 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
 304 {
 305     AVFilterContext *ctx = inlink->dst;
 306     AVFilterLink *outlink = ctx->outputs[0];
 307     LenscorrectionCtx *rect = (LenscorrectionCtx*)ctx->priv;
 308     AVFrame *out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
 309     int plane;
 310
 311     if (!out) {
 312         av_frame_free(&in);
 313         return AVERROR(ENOMEM);
 314     }
 315
 316     av_frame_copy_props(out, in);
 317
 318     for (plane = 0; plane < rect->nb_planes; ++plane) {
 319         int hsub = plane == 1 || plane == 2 ? rect->hsub : 0;
 320         int vsub = plane == 1 || plane == 2 ? rect->vsub : 0;
 321         int w = AV_CEIL_RSHIFT(rect->width, hsub);
 322         int h = AV_CEIL_RSHIFT(rect->height, vsub);
 323         int xcenter = rect->cx * w;
 324         int ycenter = rect->cy * h;
 325         ThreadData td = {
 326             .in = in,
 327             .out  = out,
 328             .w  = w,
 329             .h  = h,
 330             .xcenter = xcenter,
 331             .ycenter = ycenter,
 332             .plane = plane,
 333             .depth = rect->depth,
 334             .fill_color = rect->fill_color[plane],
 335             .correction = rect->correction[plane],
 336         };
 337
 338         ctx->internal->execute(ctx, rect->filter_slice, &td, NULL, FFMIN(h, ff_filter_get_nb_threads(ctx)));
 339     }
 340
 341     av_frame_free(&in);
 342     return ff_filter_frame(outlink, out);
 343 }
 344
 345 static int process_command(AVFilterContext *ctx,
 346                            const char *cmd,
 347                            const char *arg,
 348                            char *res,
 349                            int res_len,
 350                            int flags)
 351 {
 352     int ret = ff_filter_process_command(ctx, cmd, arg, res, res_len, flags);
 353
 354     if (ret < 0)
 355         return ret;
 356
 357     return config_output(ctx->outputs[0]);
 358 }
 359
 360 static const AVFilterPad lenscorrection_inputs[] = {
 361     {
 362         .name         = "default",
 363         .type         = AVMEDIA_TYPE_VIDEO,
 364         .filter_frame = filter_frame,
 365     },
 366     { NULL }
 367 };
 368
 369 static const AVFilterPad lenscorrection_outputs[] = {
 370     {
 371         .name         = "default",
 372         .type         = AVMEDIA_TYPE_VIDEO,
 373         .config_props = config_output,
 374     },
 375     { NULL }
 376 };
 377
 378 AVFilter ff_vf_lenscorrection = {
 379     .name          = "lenscorrection",
 380     .description   = NULL_IF_CONFIG_SMALL("Rectify the image by correcting for lens distortion."),
 381     .priv_size     = sizeof(LenscorrectionCtx),
 382     .query_formats = query_formats,
 383     .inputs        = lenscorrection_inputs,
 384     .outputs       = lenscorrection_outputs,
 385     .priv_class    = &lenscorrection_class,
 386     .uninit        = uninit,
 387     .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
 388     .process_command = process_command,
 389 };