summaryrefslogtreecommitdiffstats
path: root/media/libvpx/vp9/encoder/vp9_aq_variance.c
blob: f072717f1d9cbb9d6dc2e81a922fba41d9042907 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
/*
 *  Copyright (c) 2013 The WebM project authors. All Rights Reserved.
 *
 *  Use of this source code is governed by a BSD-style license
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
 *  in the file PATENTS.  All contributing project authors may
 *  be found in the AUTHORS file in the root of the source tree.
 */

#include <math.h>

#include "vpx_ports/mem.h"

#include "vp9/encoder/vp9_aq_variance.h"

#include "vp9/common/vp9_seg_common.h"

#include "vp9/encoder/vp9_ratectrl.h"
#include "vp9/encoder/vp9_rd.h"
#include "vp9/encoder/vp9_segmentation.h"
#include "vp9/common/vp9_systemdependent.h"

#define ENERGY_MIN (-4)
#define ENERGY_MAX (1)
#define ENERGY_SPAN (ENERGY_MAX - ENERGY_MIN +  1)
#define ENERGY_IN_BOUNDS(energy)\
  assert((energy) >= ENERGY_MIN && (energy) <= ENERGY_MAX)

static const double rate_ratio[MAX_SEGMENTS] =
  {2.5, 2.0, 1.5, 1.0, 0.75, 1.0, 1.0, 1.0};
static const int segment_id[ENERGY_SPAN] = {0, 1, 1, 2, 3, 4};

#define SEGMENT_ID(i) segment_id[(i) - ENERGY_MIN]

DECLARE_ALIGNED(16, static const uint8_t, vp9_64_zeros[64]) = {0};
#if CONFIG_VP9_HIGHBITDEPTH
DECLARE_ALIGNED(16, static const uint16_t, vp9_highbd_64_zeros[64]) = {0};
#endif

unsigned int vp9_vaq_segment_id(int energy) {
  ENERGY_IN_BOUNDS(energy);
  return SEGMENT_ID(energy);
}

void vp9_vaq_frame_setup(VP9_COMP *cpi) {
  VP9_COMMON *cm = &cpi->common;
  struct segmentation *seg = &cm->seg;
  int i;

  if (cm->frame_type == KEY_FRAME ||
      cpi->refresh_alt_ref_frame ||
      (cpi->refresh_golden_frame && !cpi->rc.is_src_frame_alt_ref)) {
    vp9_enable_segmentation(seg);
    vp9_clearall_segfeatures(seg);

    seg->abs_delta = SEGMENT_DELTADATA;

    vp9_clear_system_state();

    for (i = 0; i < MAX_SEGMENTS; ++i) {
      int qindex_delta =
          vp9_compute_qdelta_by_rate(&cpi->rc, cm->frame_type, cm->base_qindex,
                                     rate_ratio[i], cm->bit_depth);

      // We don't allow qindex 0 in a segment if the base value is not 0.
      // Q index 0 (lossless) implies 4x4 encoding only and in AQ mode a segment
      // Q delta is sometimes applied without going back around the rd loop.
      // This could lead to an illegal combination of partition size and q.
      if ((cm->base_qindex != 0) && ((cm->base_qindex + qindex_delta) == 0)) {
        qindex_delta = -cm->base_qindex + 1;
      }

      // No need to enable SEG_LVL_ALT_Q for this segment.
      if (rate_ratio[i] == 1.0) {
        continue;
      }

      vp9_set_segdata(seg, i, SEG_LVL_ALT_Q, qindex_delta);
      vp9_enable_segfeature(seg, i, SEG_LVL_ALT_Q);
    }
  }
}

/* TODO(agrange, paulwilkins): The block_variance calls the unoptimized versions
 * of variance() and highbd_8_variance(). It should not.
 */
static void aq_variance(const uint8_t *a, int  a_stride,
                        const uint8_t *b, int  b_stride,
                        int  w, int  h, unsigned int *sse, int *sum) {
  int i, j;

  *sum = 0;
  *sse = 0;

  for (i = 0; i < h; i++) {
    for (j = 0; j < w; j++) {
      const int diff = a[j] - b[j];
      *sum += diff;
      *sse += diff * diff;
    }

    a += a_stride;
    b += b_stride;
  }
}

#if CONFIG_VP9_HIGHBITDEPTH
static void aq_highbd_variance64(const uint8_t *a8, int  a_stride,
                                 const uint8_t *b8, int  b_stride,
                                 int w, int h, uint64_t *sse, uint64_t *sum) {
  int i, j;

  uint16_t *a = CONVERT_TO_SHORTPTR(a8);
  uint16_t *b = CONVERT_TO_SHORTPTR(b8);
  *sum = 0;
  *sse = 0;

  for (i = 0; i < h; i++) {
    for (j = 0; j < w; j++) {
      const int diff = a[j] - b[j];
      *sum += diff;
      *sse += diff * diff;
    }
    a += a_stride;
    b += b_stride;
  }
}

static void aq_highbd_8_variance(const uint8_t *a8, int  a_stride,
                                 const uint8_t *b8, int  b_stride,
                                 int w, int h, unsigned int *sse, int *sum) {
  uint64_t sse_long = 0;
  uint64_t sum_long = 0;
  aq_highbd_variance64(a8, a_stride, b8, b_stride, w, h, &sse_long, &sum_long);
  *sse = (unsigned int)sse_long;
  *sum = (int)sum_long;
}
#endif  // CONFIG_VP9_HIGHBITDEPTH

static unsigned int block_variance(VP9_COMP *cpi, MACROBLOCK *x,
                                   BLOCK_SIZE bs) {
  MACROBLOCKD *xd = &x->e_mbd;
  unsigned int var, sse;
  int right_overflow = (xd->mb_to_right_edge < 0) ?
      ((-xd->mb_to_right_edge) >> 3) : 0;
  int bottom_overflow = (xd->mb_to_bottom_edge < 0) ?
      ((-xd->mb_to_bottom_edge) >> 3) : 0;

  if (right_overflow || bottom_overflow) {
    const int bw = 8 * num_8x8_blocks_wide_lookup[bs] - right_overflow;
    const int bh = 8 * num_8x8_blocks_high_lookup[bs] - bottom_overflow;
    int avg;
#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
      aq_highbd_8_variance(x->plane[0].src.buf, x->plane[0].src.stride,
                           CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros), 0, bw, bh,
                           &sse, &avg);
      sse >>= 2 * (xd->bd - 8);
      avg >>= (xd->bd - 8);
    } else {
      aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
                  vp9_64_zeros, 0, bw, bh, &sse, &avg);
    }
#else
    aq_variance(x->plane[0].src.buf, x->plane[0].src.stride,
                vp9_64_zeros, 0, bw, bh, &sse, &avg);
#endif  // CONFIG_VP9_HIGHBITDEPTH
    var = sse - (((int64_t)avg * avg) / (bw * bh));
    return (256 * var) / (bw * bh);
  } else {
#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
      var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
                               x->plane[0].src.stride,
                               CONVERT_TO_BYTEPTR(vp9_highbd_64_zeros),
                               0, &sse);
    } else {
      var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
                               x->plane[0].src.stride,
                               vp9_64_zeros, 0, &sse);
    }
#else
    var = cpi->fn_ptr[bs].vf(x->plane[0].src.buf,
                             x->plane[0].src.stride,
                             vp9_64_zeros, 0, &sse);
#endif  // CONFIG_VP9_HIGHBITDEPTH
    return (256 * var) >> num_pels_log2_lookup[bs];
  }
}

double vp9_log_block_var(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) {
  unsigned int var = block_variance(cpi, x, bs);
  vp9_clear_system_state();
  return log(var + 1.0);
}

#define DEFAULT_E_MIDPOINT 10.0
int vp9_block_energy(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) {
  double energy;
  double energy_midpoint;
  vp9_clear_system_state();
  energy_midpoint =
    (cpi->oxcf.pass == 2) ? cpi->twopass.mb_av_energy : DEFAULT_E_MIDPOINT;
  energy = vp9_log_block_var(cpi, x, bs) - energy_midpoint;
  return clamp((int)round(energy), ENERGY_MIN, ENERGY_MAX);
}