pickrst.c 42.8 KB
Newer Older
1
/*
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3
 *
4 5 6 7 8 9
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 11 12
 */

#include <assert.h>
13
#include <float.h>
14 15 16
#include <limits.h>
#include <math.h>

Yaowu Xu's avatar
Yaowu Xu committed
17
#include "./aom_scale_rtcd.h"
18

19
#include "aom_dsp/psnr.h"
Yaowu Xu's avatar
Yaowu Xu committed
20 21
#include "aom_dsp/aom_dsp_common.h"
#include "aom_mem/aom_mem.h"
22
#include "aom_ports/mem.h"
23

24 25
#include "av1/common/onyxc_int.h"
#include "av1/common/quant_common.h"
26
#include "av1/common/restoration.h"
27

28
#include "av1/encoder/av1_quantize.h"
29 30 31
#include "av1/encoder/encoder.h"
#include "av1/encoder/picklpf.h"
#include "av1/encoder/pickrst.h"
32

33
typedef double (*search_restore_type)(const YV12_BUFFER_CONFIG *src,
34 35
                                      AV1_COMP *cpi, int partial_frame,
                                      RestorationInfo *info,
36
                                      RestorationType *rest_level,
37 38
                                      double *best_tile_cost,
                                      YV12_BUFFER_CONFIG *dst_frame);
39

40
const int frame_level_restore_bits[RESTORE_TYPES] = { 2, 2, 2, 2 };
41 42

static int64_t sse_restoration_tile(const YV12_BUFFER_CONFIG *src,
43 44
                                    const YV12_BUFFER_CONFIG *dst,
                                    const AV1_COMMON *cm, int h_start,
45
                                    int width, int v_start, int height,
46 47
                                    int components_pattern) {
  int64_t filt_err = 0;
48 49 50 51
  (void)cm;
  // Y and UV components cannot be mixed
  assert(components_pattern == 1 || components_pattern == 2 ||
         components_pattern == 4 || components_pattern == 6);
52 53
#if CONFIG_AOM_HIGHBITDEPTH
  if (cm->use_highbitdepth) {
54 55 56 57 58
    if ((components_pattern >> AOM_PLANE_Y) & 1) {
      filt_err +=
          aom_highbd_get_y_sse_part(src, dst, h_start, width, v_start, height);
    }
    if ((components_pattern >> AOM_PLANE_U) & 1) {
59 60
      filt_err +=
          aom_highbd_get_u_sse_part(src, dst, h_start, width, v_start, height);
61 62
    }
    if ((components_pattern >> AOM_PLANE_V) & 1) {
63 64
      filt_err +=
          aom_highbd_get_v_sse_part(src, dst, h_start, width, v_start, height);
65 66
    }
    return filt_err;
67 68
  }
#endif  // CONFIG_AOM_HIGHBITDEPTH
69 70 71 72
  if ((components_pattern >> AOM_PLANE_Y) & 1) {
    filt_err += aom_get_y_sse_part(src, dst, h_start, width, v_start, height);
  }
  if ((components_pattern >> AOM_PLANE_U) & 1) {
73
    filt_err += aom_get_u_sse_part(src, dst, h_start, width, v_start, height);
74 75
  }
  if ((components_pattern >> AOM_PLANE_V) & 1) {
76
    filt_err += aom_get_v_sse_part(src, dst, h_start, width, v_start, height);
77
  }
78 79 80
  return filt_err;
}

81 82
static int64_t sse_restoration_frame(AV1_COMMON *const cm,
                                     const YV12_BUFFER_CONFIG *src,
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98
                                     const YV12_BUFFER_CONFIG *dst,
                                     int components_pattern) {
  int64_t filt_err = 0;
#if CONFIG_AOM_HIGHBITDEPTH
  if (cm->use_highbitdepth) {
    if ((components_pattern >> AOM_PLANE_Y) & 1) {
      filt_err += aom_highbd_get_y_sse(src, dst);
    }
    if ((components_pattern >> AOM_PLANE_U) & 1) {
      filt_err += aom_highbd_get_u_sse(src, dst);
    }
    if ((components_pattern >> AOM_PLANE_V) & 1) {
      filt_err += aom_highbd_get_v_sse(src, dst);
    }
    return filt_err;
  }
99 100
#else
  (void)cm;
101 102 103 104 105 106 107 108 109 110 111 112 113
#endif  // CONFIG_AOM_HIGHBITDEPTH
  if ((components_pattern >> AOM_PLANE_Y) & 1) {
    filt_err = aom_get_y_sse(src, dst);
  }
  if ((components_pattern >> AOM_PLANE_U) & 1) {
    filt_err += aom_get_u_sse(src, dst);
  }
  if ((components_pattern >> AOM_PLANE_V) & 1) {
    filt_err += aom_get_v_sse(src, dst);
  }
  return filt_err;
}

114 115
static int64_t try_restoration_tile(const YV12_BUFFER_CONFIG *src,
                                    AV1_COMP *const cpi, RestorationInfo *rsi,
116 117 118
                                    int components_pattern, int partial_frame,
                                    int tile_idx, int subtile_idx,
                                    int subtile_bits,
119
                                    YV12_BUFFER_CONFIG *dst_frame) {
120 121 122 123
  AV1_COMMON *const cm = &cpi->common;
  int64_t filt_err;
  int tile_width, tile_height, nhtiles, nvtiles;
  int h_start, h_end, v_start, v_end;
124 125 126 127 128 129 130 131 132 133 134 135 136
  int ntiles, width, height;

  // Y and UV components cannot be mixed
  assert(components_pattern == 1 || components_pattern == 2 ||
         components_pattern == 4 || components_pattern == 6);

  if (components_pattern == 1) {  // Y only
    width = src->y_crop_width;
    height = src->y_crop_height;
  } else {  // Color
    width = src->uv_crop_width;
    height = src->uv_crop_height;
  }
137 138 139
  ntiles = av1_get_rest_ntiles(
      width, height, cm->rst_info[components_pattern > 1].restoration_tilesize,
      &tile_width, &tile_height, &nhtiles, &nvtiles);
140 141
  (void)ntiles;

142 143
  av1_loop_restoration_frame(cm->frame_to_show, cm, rsi, components_pattern,
                             partial_frame, dst_frame);
144
  av1_get_rest_tile_limits(tile_idx, subtile_idx, subtile_bits, nhtiles,
145 146
                           nvtiles, tile_width, tile_height, width, height, 0,
                           0, &h_start, &h_end, &v_start, &v_end);
147
  filt_err = sse_restoration_tile(src, dst_frame, cm, h_start, h_end - h_start,
148
                                  v_start, v_end - v_start, components_pattern);
149 150 151 152 153

  return filt_err;
}

static int64_t try_restoration_frame(const YV12_BUFFER_CONFIG *src,
Yaowu Xu's avatar
Yaowu Xu committed
154
                                     AV1_COMP *const cpi, RestorationInfo *rsi,
155
                                     int components_pattern, int partial_frame,
156
                                     YV12_BUFFER_CONFIG *dst_frame) {
Yaowu Xu's avatar
Yaowu Xu committed
157
  AV1_COMMON *const cm = &cpi->common;
158
  int64_t filt_err;
159 160
  av1_loop_restoration_frame(cm->frame_to_show, cm, rsi, components_pattern,
                             partial_frame, dst_frame);
161
  filt_err = sse_restoration_frame(cm, src, dst_frame, components_pattern);
162 163 164
  return filt_err;
}

165 166 167 168 169
static int64_t get_pixel_proj_error(uint8_t *src8, int width, int height,
                                    int src_stride, uint8_t *dat8,
                                    int dat_stride, int bit_depth,
                                    int32_t *flt1, int flt1_stride,
                                    int32_t *flt2, int flt2_stride, int *xqd) {
170 171 172 173
  int i, j;
  int64_t err = 0;
  int xq[2];
  decode_xq(xqd, xq);
174 175 176 177 178 179 180 181 182
  if (bit_depth == 8) {
    const uint8_t *src = src8;
    const uint8_t *dat = dat8;
    for (i = 0; i < height; ++i) {
      for (j = 0; j < width; ++j) {
        const int32_t u =
            (int32_t)(dat[i * dat_stride + j] << SGRPROJ_RST_BITS);
        const int32_t f1 = (int32_t)flt1[i * flt1_stride + j] - u;
        const int32_t f2 = (int32_t)flt2[i * flt2_stride + j] - u;
David Barker's avatar
David Barker committed
183
        const int32_t v = xq[0] * f1 + xq[1] * f2 + (u << SGRPROJ_PRJ_BITS);
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
        const int32_t e =
            ROUND_POWER_OF_TWO(v, SGRPROJ_RST_BITS + SGRPROJ_PRJ_BITS) -
            src[i * src_stride + j];
        err += e * e;
      }
    }
  } else {
    const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
    const uint16_t *dat = CONVERT_TO_SHORTPTR(dat8);
    for (i = 0; i < height; ++i) {
      for (j = 0; j < width; ++j) {
        const int32_t u =
            (int32_t)(dat[i * dat_stride + j] << SGRPROJ_RST_BITS);
        const int32_t f1 = (int32_t)flt1[i * flt1_stride + j] - u;
        const int32_t f2 = (int32_t)flt2[i * flt2_stride + j] - u;
David Barker's avatar
David Barker committed
199
        const int32_t v = xq[0] * f1 + xq[1] * f2 + (u << SGRPROJ_PRJ_BITS);
200 201 202 203 204
        const int32_t e =
            ROUND_POWER_OF_TWO(v, SGRPROJ_RST_BITS + SGRPROJ_PRJ_BITS) -
            src[i * src_stride + j];
        err += e * e;
      }
205 206 207 208 209
    }
  }
  return err;
}

210 211 212 213
static void get_proj_subspace(uint8_t *src8, int width, int height,
                              int src_stride, uint8_t *dat8, int dat_stride,
                              int bit_depth, int32_t *flt1, int flt1_stride,
                              int32_t *flt2, int flt2_stride, int *xq) {
214 215 216 217 218 219 220
  int i, j;
  double H[2][2] = { { 0, 0 }, { 0, 0 } };
  double C[2] = { 0, 0 };
  double Det;
  double x[2];
  const int size = width * height;

221 222 223
  // Default
  xq[0] = 0;
  xq[1] = 0;
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
  if (bit_depth == 8) {
    const uint8_t *src = src8;
    const uint8_t *dat = dat8;
    for (i = 0; i < height; ++i) {
      for (j = 0; j < width; ++j) {
        const double u = (double)(dat[i * dat_stride + j] << SGRPROJ_RST_BITS);
        const double s =
            (double)(src[i * src_stride + j] << SGRPROJ_RST_BITS) - u;
        const double f1 = (double)flt1[i * flt1_stride + j] - u;
        const double f2 = (double)flt2[i * flt2_stride + j] - u;
        H[0][0] += f1 * f1;
        H[1][1] += f2 * f2;
        H[0][1] += f1 * f2;
        C[0] += f1 * s;
        C[1] += f2 * s;
      }
    }
  } else {
    const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
    const uint16_t *dat = CONVERT_TO_SHORTPTR(dat8);
    for (i = 0; i < height; ++i) {
      for (j = 0; j < width; ++j) {
        const double u = (double)(dat[i * dat_stride + j] << SGRPROJ_RST_BITS);
        const double s =
            (double)(src[i * src_stride + j] << SGRPROJ_RST_BITS) - u;
        const double f1 = (double)flt1[i * flt1_stride + j] - u;
        const double f2 = (double)flt2[i * flt2_stride + j] - u;
        H[0][0] += f1 * f1;
        H[1][1] += f2 * f2;
        H[0][1] += f1 * f2;
        C[0] += f1 * s;
        C[1] += f2 * s;
      }
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
    }
  }
  H[0][0] /= size;
  H[0][1] /= size;
  H[1][1] /= size;
  H[1][0] = H[0][1];
  C[0] /= size;
  C[1] /= size;
  Det = (H[0][0] * H[1][1] - H[0][1] * H[1][0]);
  if (Det < 1e-8) return;  // ill-posed, return default values
  x[0] = (H[1][1] * C[0] - H[0][1] * C[1]) / Det;
  x[1] = (H[0][0] * C[1] - H[1][0] * C[0]) / Det;
  xq[0] = (int)rint(x[0] * (1 << SGRPROJ_PRJ_BITS));
  xq[1] = (int)rint(x[1] * (1 << SGRPROJ_PRJ_BITS));
}

void encode_xq(int *xq, int *xqd) {
274
  xqd[0] = xq[0];
275
  xqd[0] = clamp(xqd[0], SGRPROJ_PRJ_MIN0, SGRPROJ_PRJ_MAX0);
276
  xqd[1] = (1 << SGRPROJ_PRJ_BITS) - xqd[0] - xq[1];
277 278 279 280 281 282
  xqd[1] = clamp(xqd[1], SGRPROJ_PRJ_MIN1, SGRPROJ_PRJ_MAX1);
}

static void search_selfguided_restoration(uint8_t *dat8, int width, int height,
                                          int dat_stride, uint8_t *src8,
                                          int src_stride, int bit_depth,
283 284
                                          int *eps, int *xqd, int32_t *rstbuf) {
  int32_t *flt1 = rstbuf;
285
  int32_t *flt2 = flt1 + RESTORATION_TILEPELS_MAX;
286
  int32_t *tmpbuf2 = flt2 + RESTORATION_TILEPELS_MAX;
287
  int ep, bestep = 0;
288 289
  int64_t err, besterr = -1;
  int exqd[2], bestxqd[2] = { 0, 0 };
290

291 292
  for (ep = 0; ep < SGRPROJ_PARAMS; ep++) {
    int exq[2];
293
#if CONFIG_AOM_HIGHBITDEPTH
294 295
    if (bit_depth > 8) {
      uint16_t *dat = CONVERT_TO_SHORTPTR(dat8);
296 297
#if USE_HIGHPASS_IN_SGRPROJ
      av1_highpass_filter_highbd(dat, width, height, dat_stride, flt1, width,
298
                                 sgr_params[ep].corner, sgr_params[ep].edge);
299
#else
300 301 302
      av1_selfguided_restoration_highbd(dat, width, height, dat_stride, flt1,
                                        width, bit_depth, sgr_params[ep].r1,
                                        sgr_params[ep].e1, tmpbuf2);
303
#endif  // USE_HIGHPASS_IN_SGRPROJ
304 305 306
      av1_selfguided_restoration_highbd(dat, width, height, dat_stride, flt2,
                                        width, bit_depth, sgr_params[ep].r2,
                                        sgr_params[ep].e2, tmpbuf2);
307
    } else {
308
#endif
309 310
#if USE_HIGHPASS_IN_SGRPROJ
      av1_highpass_filter(dat8, width, height, dat_stride, flt1, width,
311
                          sgr_params[ep].corner, sgr_params[ep].edge);
312 313 314 315
#else
    av1_selfguided_restoration(dat8, width, height, dat_stride, flt1, width,
                               sgr_params[ep].r1, sgr_params[ep].e1, tmpbuf2);
#endif  // USE_HIGHPASS_IN_SGRPROJ
316
      av1_selfguided_restoration(dat8, width, height, dat_stride, flt2, width,
317
                                 sgr_params[ep].r2, sgr_params[ep].e2, tmpbuf2);
318
#if CONFIG_AOM_HIGHBITDEPTH
319
    }
320
#endif
321 322
    get_proj_subspace(src8, width, height, src_stride, dat8, dat_stride,
                      bit_depth, flt1, width, flt2, width, exq);
323
    encode_xq(exq, exqd);
324 325 326
    err =
        get_pixel_proj_error(src8, width, height, src_stride, dat8, dat_stride,
                             bit_depth, flt1, width, flt2, width, exqd);
327 328 329 330 331 332 333 334 335 336 337 338 339
    if (besterr == -1 || err < besterr) {
      bestep = ep;
      besterr = err;
      bestxqd[0] = exqd[0];
      bestxqd[1] = exqd[1];
    }
  }
  *eps = bestep;
  xqd[0] = bestxqd[0];
  xqd[1] = bestxqd[1];
}

static double search_sgrproj(const YV12_BUFFER_CONFIG *src, AV1_COMP *cpi,
340 341
                             int partial_frame, RestorationInfo *info,
                             RestorationType *type, double *best_tile_cost,
342
                             YV12_BUFFER_CONFIG *dst_frame) {
343 344 345 346 347 348
  SgrprojInfo *sgrproj_info = info->sgrproj_info;
  double err, cost_norestore, cost_sgrproj;
  int bits;
  MACROBLOCK *x = &cpi->td.mb;
  AV1_COMMON *const cm = &cpi->common;
  const YV12_BUFFER_CONFIG *dgd = cm->frame_to_show;
349
  RestorationInfo *rsi = &cpi->rst_search[0];
350 351
  int tile_idx, tile_width, tile_height, nhtiles, nvtiles;
  int h_start, h_end, v_start, v_end;
352
  // Allocate for the src buffer at high precision
353 354 355
  const int ntiles = av1_get_rest_ntiles(
      cm->width, cm->height, cm->rst_info[0].restoration_tilesize, &tile_width,
      &tile_height, &nhtiles, &nvtiles);
356
  rsi->frame_restoration_type = RESTORE_SGRPROJ;
357

358 359 360
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    rsi->restoration_type[tile_idx] = RESTORE_NONE;
  }
361 362 363 364 365
  // Compute best Sgrproj filters for each tile
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    av1_get_rest_tile_limits(tile_idx, 0, 0, nhtiles, nvtiles, tile_width,
                             tile_height, cm->width, cm->height, 0, 0, &h_start,
                             &h_end, &v_start, &v_end);
366
    err = sse_restoration_tile(src, cm->frame_to_show, cm, h_start,
367
                               h_end - h_start, v_start, v_end - v_start, 1);
368 369 370 371 372 373 374 375 376 377 378 379 380
    // #bits when a tile is not restored
    bits = av1_cost_bit(RESTORE_NONE_SGRPROJ_PROB, 0);
    cost_norestore = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);
    best_tile_cost[tile_idx] = DBL_MAX;
    search_selfguided_restoration(
        dgd->y_buffer + v_start * dgd->y_stride + h_start, h_end - h_start,
        v_end - v_start, dgd->y_stride,
        src->y_buffer + v_start * src->y_stride + h_start, src->y_stride,
#if CONFIG_AOM_HIGHBITDEPTH
        cm->bit_depth,
#else
        8,
#endif  // CONFIG_AOM_HIGHBITDEPTH
381
        &rsi->sgrproj_info[tile_idx].ep, rsi->sgrproj_info[tile_idx].xqd,
382
        cm->rst_internal.tmpbuf);
383
    rsi->restoration_type[tile_idx] = RESTORE_SGRPROJ;
384
    err = try_restoration_tile(src, cpi, rsi, 1, partial_frame, tile_idx, 0, 0,
385
                               dst_frame);
386 387 388 389
    bits = SGRPROJ_BITS << AV1_PROB_COST_SHIFT;
    bits += av1_cost_bit(RESTORE_NONE_SGRPROJ_PROB, 1);
    cost_sgrproj = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);
    if (cost_sgrproj >= cost_norestore) {
390
      type[tile_idx] = RESTORE_NONE;
391
    } else {
392
      type[tile_idx] = RESTORE_SGRPROJ;
393
      memcpy(&sgrproj_info[tile_idx], &rsi->sgrproj_info[tile_idx],
394 395 396 397 398 399
             sizeof(sgrproj_info[tile_idx]));
      bits = SGRPROJ_BITS << AV1_PROB_COST_SHIFT;
      best_tile_cost[tile_idx] = RDCOST_DBL(
          x->rdmult, x->rddiv,
          (bits + cpi->switchable_restore_cost[RESTORE_SGRPROJ]) >> 4, err);
    }
400
    rsi->restoration_type[tile_idx] = RESTORE_NONE;
401 402
  }
  // Cost for Sgrproj filtering
403
  bits = frame_level_restore_bits[rsi->frame_restoration_type]
404 405 406
         << AV1_PROB_COST_SHIFT;
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    bits +=
407
        av1_cost_bit(RESTORE_NONE_SGRPROJ_PROB, type[tile_idx] != RESTORE_NONE);
408
    memcpy(&rsi->sgrproj_info[tile_idx], &sgrproj_info[tile_idx],
409
           sizeof(sgrproj_info[tile_idx]));
410
    if (type[tile_idx] == RESTORE_SGRPROJ) {
411 412
      bits += (SGRPROJ_BITS << AV1_PROB_COST_SHIFT);
    }
413
    rsi->restoration_type[tile_idx] = type[tile_idx];
414
  }
415
  err = try_restoration_frame(src, cpi, rsi, 1, partial_frame, dst_frame);
416 417 418 419 420
  cost_sgrproj = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);

  return cost_sgrproj;
}

421 422
static double find_average(uint8_t *src, int h_start, int h_end, int v_start,
                           int v_end, int stride) {
423 424 425
  uint64_t sum = 0;
  double avg = 0;
  int i, j;
426 427 428
  for (i = v_start; i < v_end; i++)
    for (j = h_start; j < h_end; j++) sum += src[i * stride + j];
  avg = (double)sum / ((v_end - v_start) * (h_end - h_start));
429 430 431
  return avg;
}

432 433 434
static void compute_stats(uint8_t *dgd, uint8_t *src, int h_start, int h_end,
                          int v_start, int v_end, int dgd_stride,
                          int src_stride, double *M, double *H) {
435
  int i, j, k, l;
436
  double Y[WIENER_WIN2];
437 438
  const double avg =
      find_average(dgd, h_start, h_end, v_start, v_end, dgd_stride);
439

440 441
  memset(M, 0, sizeof(*M) * WIENER_WIN2);
  memset(H, 0, sizeof(*H) * WIENER_WIN2 * WIENER_WIN2);
442 443
  for (i = v_start; i < v_end; i++) {
    for (j = h_start; j < h_end; j++) {
444 445
      const double X = (double)src[i * src_stride + j] - avg;
      int idx = 0;
446 447
      for (k = -WIENER_HALFWIN; k <= WIENER_HALFWIN; k++) {
        for (l = -WIENER_HALFWIN; l <= WIENER_HALFWIN; l++) {
448 449 450 451
          Y[idx] = (double)dgd[(i + l) * dgd_stride + (j + k)] - avg;
          idx++;
        }
      }
452
      for (k = 0; k < WIENER_WIN2; ++k) {
453
        M[k] += Y[k] * X;
454 455
        H[k * WIENER_WIN2 + k] += Y[k] * Y[k];
        for (l = k + 1; l < WIENER_WIN2; ++l) {
456 457 458 459
          // H is a symmetric matrix, so we only need to fill out the upper
          // triangle here. We can copy it down to the lower triangle outside
          // the (i, j) loops.
          H[k * WIENER_WIN2 + l] += Y[k] * Y[l];
460 461 462 463
        }
      }
    }
  }
464 465 466 467 468
  for (k = 0; k < WIENER_WIN2; ++k) {
    for (l = k + 1; l < WIENER_WIN2; ++l) {
      H[l * WIENER_WIN2 + k] = H[k * WIENER_WIN2 + l];
    }
  }
469 470
}

Yaowu Xu's avatar
Yaowu Xu committed
471
#if CONFIG_AOM_HIGHBITDEPTH
472 473
static double find_average_highbd(uint16_t *src, int h_start, int h_end,
                                  int v_start, int v_end, int stride) {
474 475 476
  uint64_t sum = 0;
  double avg = 0;
  int i, j;
477 478 479
  for (i = v_start; i < v_end; i++)
    for (j = h_start; j < h_end; j++) sum += src[i * stride + j];
  avg = (double)sum / ((v_end - v_start) * (h_end - h_start));
480 481 482
  return avg;
}

483 484 485 486
static void compute_stats_highbd(uint8_t *dgd8, uint8_t *src8, int h_start,
                                 int h_end, int v_start, int v_end,
                                 int dgd_stride, int src_stride, double *M,
                                 double *H) {
487
  int i, j, k, l;
488
  double Y[WIENER_WIN2];
489 490
  uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  uint16_t *dgd = CONVERT_TO_SHORTPTR(dgd8);
491 492
  const double avg =
      find_average_highbd(dgd, h_start, h_end, v_start, v_end, dgd_stride);
493

494 495
  memset(M, 0, sizeof(*M) * WIENER_WIN2);
  memset(H, 0, sizeof(*H) * WIENER_WIN2 * WIENER_WIN2);
496 497
  for (i = v_start; i < v_end; i++) {
    for (j = h_start; j < h_end; j++) {
498 499
      const double X = (double)src[i * src_stride + j] - avg;
      int idx = 0;
500 501
      for (k = -WIENER_HALFWIN; k <= WIENER_HALFWIN; k++) {
        for (l = -WIENER_HALFWIN; l <= WIENER_HALFWIN; l++) {
502 503 504 505
          Y[idx] = (double)dgd[(i + l) * dgd_stride + (j + k)] - avg;
          idx++;
        }
      }
506
      for (k = 0; k < WIENER_WIN2; ++k) {
507
        M[k] += Y[k] * X;
508 509
        H[k * WIENER_WIN2 + k] += Y[k] * Y[k];
        for (l = k + 1; l < WIENER_WIN2; ++l) {
510 511 512 513
          // H is a symmetric matrix, so we only need to fill out the upper
          // triangle here. We can copy it down to the lower triangle outside
          // the (i, j) loops.
          H[k * WIENER_WIN2 + l] += Y[k] * Y[l];
514 515 516 517
        }
      }
    }
  }
518 519 520 521 522
  for (k = 0; k < WIENER_WIN2; ++k) {
    for (l = k + 1; l < WIENER_WIN2; ++l) {
      H[l * WIENER_WIN2 + k] = H[k * WIENER_WIN2 + l];
    }
  }
523
}
Yaowu Xu's avatar
Yaowu Xu committed
524
#endif  // CONFIG_AOM_HIGHBITDEPTH
525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546

// Solves Ax = b, where x and b are column vectors
static int linsolve(int n, double *A, int stride, double *b, double *x) {
  int i, j, k;
  double c;
  // Partial pivoting
  for (i = n - 1; i > 0; i--) {
    if (A[(i - 1) * stride] < A[i * stride]) {
      for (j = 0; j < n; j++) {
        c = A[i * stride + j];
        A[i * stride + j] = A[(i - 1) * stride + j];
        A[(i - 1) * stride + j] = c;
      }
      c = b[i];
      b[i] = b[i - 1];
      b[i - 1] = c;
    }
  }
  // Forward elimination
  for (k = 0; k < n - 1; k++) {
    for (i = k; i < n - 1; i++) {
      c = A[(i + 1) * stride + k] / A[k * stride + k];
547
      for (j = 0; j < n; j++) A[(i + 1) * stride + j] -= c * A[k * stride + j];
548 549 550 551 552
      b[i + 1] -= c * b[k];
    }
  }
  // Backward substitution
  for (i = n - 1; i >= 0; i--) {
553
    if (fabs(A[i * stride + i]) < 1e-10) return 0;
554
    c = 0;
555
    for (j = i + 1; j <= n - 1; j++) c += A[i * stride + j] * x[j];
556 557 558 559 560 561
    x[i] = (b[i] - c) / A[i * stride + i];
  }
  return 1;
}

static INLINE int wrap_index(int i) {
562
  return (i >= WIENER_HALFWIN1 ? WIENER_WIN - 1 - i : i);
563 564 565 566 567
}

// Fix vector b, update vector a
static void update_a_sep_sym(double **Mc, double **Hc, double *a, double *b) {
  int i, j;
568 569
  double S[WIENER_WIN];
  double A[WIENER_WIN], B[WIENER_WIN2];
Aamir Anis's avatar
Aamir Anis committed
570
  int w, w2;
571 572
  memset(A, 0, sizeof(A));
  memset(B, 0, sizeof(B));
573 574
  for (i = 0; i < WIENER_WIN; i++) {
    for (j = 0; j < WIENER_WIN; ++j) {
575 576 577 578
      const int jj = wrap_index(j);
      A[jj] += Mc[i][j] * b[i];
    }
  }
579 580
  for (i = 0; i < WIENER_WIN; i++) {
    for (j = 0; j < WIENER_WIN; j++) {
581
      int k, l;
582 583
      for (k = 0; k < WIENER_WIN; ++k)
        for (l = 0; l < WIENER_WIN; ++l) {
584 585
          const int kk = wrap_index(k);
          const int ll = wrap_index(l);
586 587
          B[ll * WIENER_HALFWIN1 + kk] +=
              Hc[j * WIENER_WIN + i][k * WIENER_WIN2 + l] * b[i] * b[j];
588 589 590
        }
    }
  }
Aamir Anis's avatar
Aamir Anis committed
591
  // Normalization enforcement in the system of equations itself
592
  w = WIENER_WIN;
Aamir Anis's avatar
Aamir Anis committed
593 594
  w2 = (w >> 1) + 1;
  for (i = 0; i < w2 - 1; ++i)
595 596
    A[i] -=
        A[w2 - 1] * 2 + B[i * w2 + w2 - 1] - 2 * B[(w2 - 1) * w2 + (w2 - 1)];
Aamir Anis's avatar
Aamir Anis committed
597 598 599 600 601 602 603 604 605
  for (i = 0; i < w2 - 1; ++i)
    for (j = 0; j < w2 - 1; ++j)
      B[i * w2 + j] -= 2 * (B[i * w2 + (w2 - 1)] + B[(w2 - 1) * w2 + j] -
                            2 * B[(w2 - 1) * w2 + (w2 - 1)]);
  if (linsolve(w2 - 1, B, w2, A, S)) {
    S[w2 - 1] = 1.0;
    for (i = w2; i < w; ++i) {
      S[i] = S[w - 1 - i];
      S[w2 - 1] -= 2 * S[i];
606
    }
Aamir Anis's avatar
Aamir Anis committed
607
    memcpy(a, S, w * sizeof(*a));
608 609 610 611 612 613
  }
}

// Fix vector a, update vector b
static void update_b_sep_sym(double **Mc, double **Hc, double *a, double *b) {
  int i, j;
614 615
  double S[WIENER_WIN];
  double A[WIENER_WIN], B[WIENER_WIN2];
Aamir Anis's avatar
Aamir Anis committed
616
  int w, w2;
617 618
  memset(A, 0, sizeof(A));
  memset(B, 0, sizeof(B));
619
  for (i = 0; i < WIENER_WIN; i++) {
620
    const int ii = wrap_index(i);
621
    for (j = 0; j < WIENER_WIN; j++) A[ii] += Mc[i][j] * a[j];
622 623
  }

624 625
  for (i = 0; i < WIENER_WIN; i++) {
    for (j = 0; j < WIENER_WIN; j++) {
626 627 628
      const int ii = wrap_index(i);
      const int jj = wrap_index(j);
      int k, l;
629 630 631 632
      for (k = 0; k < WIENER_WIN; ++k)
        for (l = 0; l < WIENER_WIN; ++l)
          B[jj * WIENER_HALFWIN1 + ii] +=
              Hc[i * WIENER_WIN + j][k * WIENER_WIN2 + l] * a[k] * a[l];
633 634
    }
  }
Aamir Anis's avatar
Aamir Anis committed
635
  // Normalization enforcement in the system of equations itself
636 637
  w = WIENER_WIN;
  w2 = WIENER_HALFWIN1;
Aamir Anis's avatar
Aamir Anis committed
638
  for (i = 0; i < w2 - 1; ++i)
639 640
    A[i] -=
        A[w2 - 1] * 2 + B[i * w2 + w2 - 1] - 2 * B[(w2 - 1) * w2 + (w2 - 1)];
Aamir Anis's avatar
Aamir Anis committed
641 642 643 644 645 646 647 648 649
  for (i = 0; i < w2 - 1; ++i)
    for (j = 0; j < w2 - 1; ++j)
      B[i * w2 + j] -= 2 * (B[i * w2 + (w2 - 1)] + B[(w2 - 1) * w2 + j] -
                            2 * B[(w2 - 1) * w2 + (w2 - 1)]);
  if (linsolve(w2 - 1, B, w2, A, S)) {
    S[w2 - 1] = 1.0;
    for (i = w2; i < w; ++i) {
      S[i] = S[w - 1 - i];
      S[w2 - 1] -= 2 * S[i];
650
    }
Aamir Anis's avatar
Aamir Anis committed
651
    memcpy(b, S, w * sizeof(*b));
652 653 654
  }
}

655 656
static int wiener_decompose_sep_sym(double *M, double *H, double *a,
                                    double *b) {
657
  static const double init_filt[WIENER_WIN] = {
658
    0.035623, -0.127154, 0.211436, 0.760190, 0.211436, -0.127154, 0.035623,
659 660
  };
  int i, j, iter;
661 662 663 664 665 666 667
  double *Hc[WIENER_WIN2];
  double *Mc[WIENER_WIN];
  for (i = 0; i < WIENER_WIN; i++) {
    Mc[i] = M + i * WIENER_WIN;
    for (j = 0; j < WIENER_WIN; j++) {
      Hc[i * WIENER_WIN + j] =
          H + i * WIENER_WIN * WIENER_WIN2 + j * WIENER_WIN;
668 669
    }
  }
670 671
  memcpy(a, init_filt, sizeof(*a) * WIENER_WIN);
  memcpy(b, init_filt, sizeof(*b) * WIENER_WIN);
672 673 674 675 676 677 678

  iter = 1;
  while (iter < 10) {
    update_a_sep_sym(Mc, Hc, a, b);
    update_b_sep_sym(Mc, Hc, a, b);
    iter++;
  }
679
  return 1;
680 681
}

682
// Computes the function x'*H*x - x'*M for the learned 2D filter x, and compares
Aamir Anis's avatar
Aamir Anis committed
683 684
// against identity filters; Final score is defined as the difference between
// the function values
685 686
static double compute_score(double *M, double *H, InterpKernel vfilt,
                            InterpKernel hfilt) {
687
  double ab[WIENER_WIN * WIENER_WIN];
Aamir Anis's avatar
Aamir Anis committed
688 689 690 691
  int i, k, l;
  double P = 0, Q = 0;
  double iP = 0, iQ = 0;
  double Score, iScore;
692 693 694 695 696 697 698
  double a[WIENER_WIN], b[WIENER_WIN];
  a[WIENER_HALFWIN] = b[WIENER_HALFWIN] = 1.0;
  for (i = 0; i < WIENER_HALFWIN; ++i) {
    a[i] = a[WIENER_WIN - i - 1] = (double)vfilt[i] / WIENER_FILT_STEP;
    b[i] = b[WIENER_WIN - i - 1] = (double)hfilt[i] / WIENER_FILT_STEP;
    a[WIENER_HALFWIN] -= 2 * a[i];
    b[WIENER_HALFWIN] -= 2 * b[i];
Aamir Anis's avatar
Aamir Anis committed
699
  }
700 701
  for (k = 0; k < WIENER_WIN; ++k) {
    for (l = 0; l < WIENER_WIN; ++l) ab[k * WIENER_WIN + l] = a[l] * b[k];
Aamir Anis's avatar
Aamir Anis committed
702
  }
703
  for (k = 0; k < WIENER_WIN2; ++k) {
Aamir Anis's avatar
Aamir Anis committed
704
    P += ab[k] * M[k];
705 706
    for (l = 0; l < WIENER_WIN2; ++l)
      Q += ab[k] * H[k * WIENER_WIN2 + l] * ab[l];
Aamir Anis's avatar
Aamir Anis committed
707 708 709
  }
  Score = Q - 2 * P;

710 711
  iP = M[WIENER_WIN2 >> 1];
  iQ = H[(WIENER_WIN2 >> 1) * WIENER_WIN2 + (WIENER_WIN2 >> 1)];
Aamir Anis's avatar
Aamir Anis committed
712 713 714 715 716
  iScore = iQ - 2 * iP;

  return Score - iScore;
}

717
static void quantize_sym_filter(double *f, InterpKernel fi) {
718
  int i;
719 720
  for (i = 0; i < WIENER_HALFWIN; ++i) {
    fi[i] = RINT(f[i] * WIENER_FILT_STEP);
721 722 723 724 725
  }
  // Specialize for 7-tap filter
  fi[0] = CLIP(fi[0], WIENER_FILT_TAP0_MINV, WIENER_FILT_TAP0_MAXV);
  fi[1] = CLIP(fi[1], WIENER_FILT_TAP1_MINV, WIENER_FILT_TAP1_MAXV);
  fi[2] = CLIP(fi[2], WIENER_FILT_TAP2_MINV, WIENER_FILT_TAP2_MAXV);
726 727 728 729
  // Satisfy filter constraints
  fi[WIENER_WIN - 1] = fi[0];
  fi[WIENER_WIN - 2] = fi[1];
  fi[WIENER_WIN - 3] = fi[2];
730 731
  // The central element has an implicit +WIENER_FILT_STEP
  fi[3] = -2 * (fi[0] + fi[1] + fi[2]);
732 733 734
}

static double search_wiener_uv(const YV12_BUFFER_CONFIG *src, AV1_COMP *cpi,
735
                               int partial_frame, int plane,
736
                               RestorationInfo *info, RestorationType *type,
737 738 739 740 741 742
                               YV12_BUFFER_CONFIG *dst_frame) {
  WienerInfo *wiener_info = info->wiener_info;
  AV1_COMMON *const cm = &cpi->common;
  RestorationInfo *rsi = cpi->rst_search;
  int64_t err;
  int bits;
743
  double cost_wiener, cost_norestore, cost_wiener_frame, cost_norestore_frame;
744 745 746 747 748 749 750 751 752 753 754
  MACROBLOCK *x = &cpi->td.mb;
  double M[WIENER_WIN2];
  double H[WIENER_WIN2 * WIENER_WIN2];
  double vfilterd[WIENER_WIN], hfilterd[WIENER_WIN];
  const YV12_BUFFER_CONFIG *dgd = cm->frame_to_show;
  const int width = src->uv_crop_width;
  const int height = src->uv_crop_height;
  const int src_stride = src->uv_stride;
  const int dgd_stride = dgd->uv_stride;
  double score;
  int tile_idx, tile_width, tile_height, nhtiles, nvtiles;
755
  int h_start, h_end, v_start, v_end;
756 757 758
  const int ntiles =
      av1_get_rest_ntiles(width, height, cm->rst_info[1].restoration_tilesize,
                          &tile_width, &tile_height, &nhtiles, &nvtiles);
759 760 761 762
  assert(width == dgd->uv_crop_width);
  assert(height == dgd->uv_crop_height);

  rsi[plane].frame_restoration_type = RESTORE_NONE;
763
  err = sse_restoration_frame(cm, src, cm->frame_to_show, (1 << plane));
764
  bits = 0;
765
  cost_norestore_frame = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);
766 767

  rsi[plane].frame_restoration_type = RESTORE_WIENER;
768

769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    rsi[plane].restoration_type[tile_idx] = RESTORE_NONE;
  }

  // Compute best Wiener filters for each tile
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    av1_get_rest_tile_limits(tile_idx, 0, 0, nhtiles, nvtiles, tile_width,
                             tile_height, width, height, 0, 0, &h_start, &h_end,
                             &v_start, &v_end);
    err = sse_restoration_tile(src, cm->frame_to_show, cm, h_start,
                               h_end - h_start, v_start, v_end - v_start,
                               1 << plane);
    // #bits when a tile is not restored
    bits = av1_cost_bit(RESTORE_NONE_WIENER_PROB, 0);
    cost_norestore = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);
    // best_tile_cost[tile_idx] = DBL_MAX;

    av1_get_rest_tile_limits(tile_idx, 0, 0, nhtiles, nvtiles, tile_width,
                             tile_height, width, height, WIENER_HALFWIN,
                             WIENER_HALFWIN, &h_start, &h_end, &v_start,
                             &v_end);
790
    if (plane == AOM_PLANE_U) {
791 792 793 794 795 796 797 798
#if CONFIG_AOM_HIGHBITDEPTH
      if (cm->use_highbitdepth)
        compute_stats_highbd(dgd->u_buffer, src->u_buffer, h_start, h_end,
                             v_start, v_end, dgd_stride, src_stride, M, H);
      else
#endif  // CONFIG_AOM_HIGHBITDEPTH
        compute_stats(dgd->u_buffer, src->u_buffer, h_start, h_end, v_start,
                      v_end, dgd_stride, src_stride, M, H);
799
    } else if (plane == AOM_PLANE_V) {
800 801 802 803 804 805 806 807
#if CONFIG_AOM_HIGHBITDEPTH
      if (cm->use_highbitdepth)
        compute_stats_highbd(dgd->v_buffer, src->v_buffer, h_start, h_end,
                             v_start, v_end, dgd_stride, src_stride, M, H);
      else
#endif  // CONFIG_AOM_HIGHBITDEPTH
        compute_stats(dgd->v_buffer, src->v_buffer, h_start, h_end, v_start,
                      v_end, dgd_stride, src_stride, M, H);
808 809 810
    } else {
      assert(0);
    }
811 812 813 814 815 816

    type[tile_idx] = RESTORE_WIENER;

    if (!wiener_decompose_sep_sym(M, H, vfilterd, hfilterd)) {
      type[tile_idx] = RESTORE_NONE;
      continue;
817
    }
818 819
    quantize_sym_filter(vfilterd, rsi[plane].wiener_info[tile_idx].vfilter);
    quantize_sym_filter(hfilterd, rsi[plane].wiener_info[tile_idx].hfilter);
820

821 822 823 824 825 826 827 828 829
    // Filter score computes the value of the function x'*A*x - x'*b for the
    // learned filter and compares it against identity filer. If there is no
    // reduction in the function, the filter is reverted back to identity
    score = compute_score(M, H, rsi[plane].wiener_info[tile_idx].vfilter,
                          rsi[plane].wiener_info[tile_idx].hfilter);
    if (score > 0.0) {
      type[tile_idx] = RESTORE_NONE;
      continue;
    }
830

831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
    rsi[plane].restoration_type[tile_idx] = RESTORE_WIENER;
    err = try_restoration_tile(src, cpi, rsi, 1 << plane, partial_frame,
                               tile_idx, 0, 0, dst_frame);
    bits = WIENER_FILT_BITS << AV1_PROB_COST_SHIFT;
    bits += av1_cost_bit(RESTORE_NONE_WIENER_PROB, 1);
    cost_wiener = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);
    if (cost_wiener >= cost_norestore) {
      type[tile_idx] = RESTORE_NONE;
    } else {
      type[tile_idx] = RESTORE_WIENER;
      memcpy(&wiener_info[tile_idx], &rsi[plane].wiener_info[tile_idx],
             sizeof(wiener_info[tile_idx]));
    }
    rsi[plane].restoration_type[tile_idx] = RESTORE_NONE;
  }
  // Cost for Wiener filtering
  bits = 0;
  for (tile_idx = 0; tile_idx < ntiles; ++tile_idx) {
    bits +=
        av1_cost_bit(RESTORE_NONE_WIENER_PROB, type[tile_idx] != RESTORE_NONE);
    memcpy(&rsi[plane].wiener_info[tile_idx], &wiener_info[tile_idx],
           sizeof(wiener_info[tile_idx]));
    if (type[tile_idx] == RESTORE_WIENER) {
      bits += (WIENER_FILT_BITS << AV1_PROB_COST_SHIFT);
    }
    rsi[plane].restoration_type[tile_idx] = type[tile_idx];
857
  }
858
  err = try_restoration_frame(src, cpi, rsi, 1 << plane, partial_frame,
859
                              dst_frame);
860 861 862 863 864
  cost_wiener_frame = RDCOST_DBL(x->rdmult, x->rddiv, (bits >> 4), err);

  if (cost_wiener_frame < cost_norestore_frame) {
    info->frame_restoration_type = RESTORE_WIENER;
  } else {
865 866 867
    info->frame_restoration_type = RESTORE_NONE;
  }

868 869
  return info->frame_restoration_type == RESTORE_WIENER ? cost_wiener_frame
                                                        : cost_norestore_frame;
870 871
}

872
static double search_wiener(const YV12_BUFFER_CONFIG *src, AV1_COMP *cpi,
873 874
                            int partial_frame, RestorationInfo *info,
                            RestorationType *type, double *best_tile_cost,
875
                            YV12_BUFFER_CONFIG *dst_frame) {
876
  WienerInfo *wiener_info = info->wiener_info;
Yaowu Xu's avatar
Yaowu Xu committed
877
  AV1_COMMON *const cm = &cpi->common;
878
  RestorationInfo *rsi = cpi->rst_search;
879 880
  int64_t err;
  int bits;
881
  double cost_wiener, cost_norestore;
882
  MACROBLOCK *x = &cpi->td.mb;
883 884 885
  double M[WIENER_WIN2];
  double H[WIENER_WIN2 * WIENER_WIN2];
  double vfilterd[WIENER_WIN], hfilterd[WIENER_WIN];
886 887 888 889 890
  const YV12_BUFFER_CONFIG *dgd = cm->frame_to_show;
  const int width = cm->width;
  const int height = cm->height;
  const int src_stride = src->y_stride;
  const int dgd_stride = dgd->y_stride;
Aamir Anis's avatar
Aamir Anis committed
891
  double score;
892
  int tile_idx, tile_width, tile_height, nhtiles, nvtiles;
893
  int h_start, h_end, v_start, v_end;
894 895 896
  const int ntiles =
      av1_get_rest_ntiles(width, height, cm->rst_info[0].restoration_tilesize,
                          &tile_width, &