/src/aom/av1/decoder/grain_synthesis.c
Line | Count | Source |
1 | | /* |
2 | | * Copyright (c) 2016, Alliance for Open Media. All rights reserved. |
3 | | * |
4 | | * This source code is subject to the terms of the BSD 2 Clause License and |
5 | | * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License |
6 | | * was not distributed with this source code in the LICENSE file, you can |
7 | | * obtain it at www.aomedia.org/license/software. If the Alliance for Open |
8 | | * Media Patent License 1.0 was not distributed with this source code in the |
9 | | * PATENTS file, you can obtain it at www.aomedia.org/license/patent. |
10 | | */ |
11 | | |
12 | | /*!\file |
13 | | * \brief Describes film grain parameters and film grain synthesis |
14 | | * |
15 | | */ |
16 | | |
17 | | #include <stdbool.h> |
18 | | #include <stdio.h> |
19 | | #include <string.h> |
20 | | #include <stdlib.h> |
21 | | #include <assert.h> |
22 | | #include "aom_dsp/aom_dsp_common.h" |
23 | | #include "aom_mem/aom_mem.h" |
24 | | #include "av1/decoder/grain_synthesis.h" |
25 | | |
26 | | // Samples with Gaussian distribution in the range of [-2048, 2047] (12 bits) |
27 | | // with zero mean and standard deviation of about 512. |
28 | | // should be divided by 4 for 10-bit range and 16 for 8-bit range. |
29 | | static const int gaussian_sequence[2048] = { |
30 | | 56, 568, -180, 172, 124, -84, 172, -64, -900, 24, 820, |
31 | | 224, 1248, 996, 272, -8, -916, -388, -732, -104, -188, 800, |
32 | | 112, -652, -320, -376, 140, -252, 492, -168, 44, -788, 588, |
33 | | -584, 500, -228, 12, 680, 272, -476, 972, -100, 652, 368, |
34 | | 432, -196, -720, -192, 1000, -332, 652, -136, -552, -604, -4, |
35 | | 192, -220, -136, 1000, -52, 372, -96, -624, 124, -24, 396, |
36 | | 540, -12, -104, 640, 464, 244, -208, -84, 368, -528, -740, |
37 | | 248, -968, -848, 608, 376, -60, -292, -40, -156, 252, -292, |
38 | | 248, 224, -280, 400, -244, 244, -60, 76, -80, 212, 532, |
39 | | 340, 128, -36, 824, -352, -60, -264, -96, -612, 416, -704, |
40 | | 220, -204, 640, -160, 1220, -408, 900, 336, 20, -336, -96, |
41 | | -792, 304, 48, -28, -1232, -1172, -448, 104, -292, -520, 244, |
42 | | 60, -948, 0, -708, 268, 108, 356, -548, 488, -344, -136, |
43 | | 488, -196, -224, 656, -236, -1128, 60, 4, 140, 276, -676, |
44 | | -376, 168, -108, 464, 8, 564, 64, 240, 308, -300, -400, |
45 | | -456, -136, 56, 120, -408, -116, 436, 504, -232, 328, 844, |
46 | | -164, -84, 784, -168, 232, -224, 348, -376, 128, 568, 96, |
47 | | -1244, -288, 276, 848, 832, -360, 656, 464, -384, -332, -356, |
48 | | 728, -388, 160, -192, 468, 296, 224, 140, -776, -100, 280, |
49 | | 4, 196, 44, -36, -648, 932, 16, 1428, 28, 528, 808, |
50 | | 772, 20, 268, 88, -332, -284, 124, -384, -448, 208, -228, |
51 | | -1044, -328, 660, 380, -148, -300, 588, 240, 540, 28, 136, |
52 | | -88, -436, 256, 296, -1000, 1400, 0, -48, 1056, -136, 264, |
53 | | -528, -1108, 632, -484, -592, -344, 796, 124, -668, -768, 388, |
54 | | 1296, -232, -188, -200, -288, -4, 308, 100, -168, 256, -500, |
55 | | 204, -508, 648, -136, 372, -272, -120, -1004, -552, -548, -384, |
56 | | 548, -296, 428, -108, -8, -912, -324, -224, -88, -112, -220, |
57 | | -100, 996, -796, 548, 360, -216, 180, 428, -200, -212, 148, |
58 | | 96, 148, 284, 216, -412, -320, 120, -300, -384, -604, -572, |
59 | | -332, -8, -180, -176, 696, 116, -88, 628, 76, 44, -516, |
60 | | 240, -208, -40, 100, -592, 344, -308, -452, -228, 20, 916, |
61 | | -1752, -136, -340, -804, 140, 40, 512, 340, 248, 184, -492, |
62 | | 896, -156, 932, -628, 328, -688, -448, -616, -752, -100, 560, |
63 | | -1020, 180, -800, -64, 76, 576, 1068, 396, 660, 552, -108, |
64 | | -28, 320, -628, 312, -92, -92, -472, 268, 16, 560, 516, |
65 | | -672, -52, 492, -100, 260, 384, 284, 292, 304, -148, 88, |
66 | | -152, 1012, 1064, -228, 164, -376, -684, 592, -392, 156, 196, |
67 | | -524, -64, -884, 160, -176, 636, 648, 404, -396, -436, 864, |
68 | | 424, -728, 988, -604, 904, -592, 296, -224, 536, -176, -920, |
69 | | 436, -48, 1176, -884, 416, -776, -824, -884, 524, -548, -564, |
70 | | -68, -164, -96, 692, 364, -692, -1012, -68, 260, -480, 876, |
71 | | -1116, 452, -332, -352, 892, -1088, 1220, -676, 12, -292, 244, |
72 | | 496, 372, -32, 280, 200, 112, -440, -96, 24, -644, -184, |
73 | | 56, -432, 224, -980, 272, -260, 144, -436, 420, 356, 364, |
74 | | -528, 76, 172, -744, -368, 404, -752, -416, 684, -688, 72, |
75 | | 540, 416, 92, 444, 480, -72, -1416, 164, -1172, -68, 24, |
76 | | 424, 264, 1040, 128, -912, -524, -356, 64, 876, -12, 4, |
77 | | -88, 532, 272, -524, 320, 276, -508, 940, 24, -400, -120, |
78 | | 756, 60, 236, -412, 100, 376, -484, 400, -100, -740, -108, |
79 | | -260, 328, -268, 224, -200, -416, 184, -604, -564, -20, 296, |
80 | | 60, 892, -888, 60, 164, 68, -760, 216, -296, 904, -336, |
81 | | -28, 404, -356, -568, -208, -1480, -512, 296, 328, -360, -164, |
82 | | -1560, -776, 1156, -428, 164, -504, -112, 120, -216, -148, -264, |
83 | | 308, 32, 64, -72, 72, 116, 176, -64, -272, 460, -536, |
84 | | -784, -280, 348, 108, -752, -132, 524, -540, -776, 116, -296, |
85 | | -1196, -288, -560, 1040, -472, 116, -848, -1116, 116, 636, 696, |
86 | | 284, -176, 1016, 204, -864, -648, -248, 356, 972, -584, -204, |
87 | | 264, 880, 528, -24, -184, 116, 448, -144, 828, 524, 212, |
88 | | -212, 52, 12, 200, 268, -488, -404, -880, 824, -672, -40, |
89 | | 908, -248, 500, 716, -576, 492, -576, 16, 720, -108, 384, |
90 | | 124, 344, 280, 576, -500, 252, 104, -308, 196, -188, -8, |
91 | | 1268, 296, 1032, -1196, 436, 316, 372, -432, -200, -660, 704, |
92 | | -224, 596, -132, 268, 32, -452, 884, 104, -1008, 424, -1348, |
93 | | -280, 4, -1168, 368, 476, 696, 300, -8, 24, 180, -592, |
94 | | -196, 388, 304, 500, 724, -160, 244, -84, 272, -256, -420, |
95 | | 320, 208, -144, -156, 156, 364, 452, 28, 540, 316, 220, |
96 | | -644, -248, 464, 72, 360, 32, -388, 496, -680, -48, 208, |
97 | | -116, -408, 60, -604, -392, 548, -840, 784, -460, 656, -544, |
98 | | -388, -264, 908, -800, -628, -612, -568, 572, -220, 164, 288, |
99 | | -16, -308, 308, -112, -636, -760, 280, -668, 432, 364, 240, |
100 | | -196, 604, 340, 384, 196, 592, -44, -500, 432, -580, -132, |
101 | | 636, -76, 392, 4, -412, 540, 508, 328, -356, -36, 16, |
102 | | -220, -64, -248, -60, 24, -192, 368, 1040, 92, -24, -1044, |
103 | | -32, 40, 104, 148, 192, -136, -520, 56, -816, -224, 732, |
104 | | 392, 356, 212, -80, -424, -1008, -324, 588, -1496, 576, 460, |
105 | | -816, -848, 56, -580, -92, -1372, -112, -496, 200, 364, 52, |
106 | | -140, 48, -48, -60, 84, 72, 40, 132, -356, -268, -104, |
107 | | -284, -404, 732, -520, 164, -304, -540, 120, 328, -76, -460, |
108 | | 756, 388, 588, 236, -436, -72, -176, -404, -316, -148, 716, |
109 | | -604, 404, -72, -88, -888, -68, 944, 88, -220, -344, 960, |
110 | | 472, 460, -232, 704, 120, 832, -228, 692, -508, 132, -476, |
111 | | 844, -748, -364, -44, 1116, -1104, -1056, 76, 428, 552, -692, |
112 | | 60, 356, 96, -384, -188, -612, -576, 736, 508, 892, 352, |
113 | | -1132, 504, -24, -352, 324, 332, -600, -312, 292, 508, -144, |
114 | | -8, 484, 48, 284, -260, -240, 256, -100, -292, -204, -44, |
115 | | 472, -204, 908, -188, -1000, -256, 92, 1164, -392, 564, 356, |
116 | | 652, -28, -884, 256, 484, -192, 760, -176, 376, -524, -452, |
117 | | -436, 860, -736, 212, 124, 504, -476, 468, 76, -472, 552, |
118 | | -692, -944, -620, 740, -240, 400, 132, 20, 192, -196, 264, |
119 | | -668, -1012, -60, 296, -316, -828, 76, -156, 284, -768, -448, |
120 | | -832, 148, 248, 652, 616, 1236, 288, -328, -400, -124, 588, |
121 | | 220, 520, -696, 1032, 768, -740, -92, -272, 296, 448, -464, |
122 | | 412, -200, 392, 440, -200, 264, -152, -260, 320, 1032, 216, |
123 | | 320, -8, -64, 156, -1016, 1084, 1172, 536, 484, -432, 132, |
124 | | 372, -52, -256, 84, 116, -352, 48, 116, 304, -384, 412, |
125 | | 924, -300, 528, 628, 180, 648, 44, -980, -220, 1320, 48, |
126 | | 332, 748, 524, -268, -720, 540, -276, 564, -344, -208, -196, |
127 | | 436, 896, 88, -392, 132, 80, -964, -288, 568, 56, -48, |
128 | | -456, 888, 8, 552, -156, -292, 948, 288, 128, -716, -292, |
129 | | 1192, -152, 876, 352, -600, -260, -812, -468, -28, -120, -32, |
130 | | -44, 1284, 496, 192, 464, 312, -76, -516, -380, -456, -1012, |
131 | | -48, 308, -156, 36, 492, -156, -808, 188, 1652, 68, -120, |
132 | | -116, 316, 160, -140, 352, 808, -416, 592, 316, -480, 56, |
133 | | 528, -204, -568, 372, -232, 752, -344, 744, -4, 324, -416, |
134 | | -600, 768, 268, -248, -88, -132, -420, -432, 80, -288, 404, |
135 | | -316, -1216, -588, 520, -108, 92, -320, 368, -480, -216, -92, |
136 | | 1688, -300, 180, 1020, -176, 820, -68, -228, -260, 436, -904, |
137 | | 20, 40, -508, 440, -736, 312, 332, 204, 760, -372, 728, |
138 | | 96, -20, -632, -520, -560, 336, 1076, -64, -532, 776, 584, |
139 | | 192, 396, -728, -520, 276, -188, 80, -52, -612, -252, -48, |
140 | | 648, 212, -688, 228, -52, -260, 428, -412, -272, -404, 180, |
141 | | 816, -796, 48, 152, 484, -88, -216, 988, 696, 188, -528, |
142 | | 648, -116, -180, 316, 476, 12, -564, 96, 476, -252, -364, |
143 | | -376, -392, 556, -256, -576, 260, -352, 120, -16, -136, -260, |
144 | | -492, 72, 556, 660, 580, 616, 772, 436, 424, -32, -324, |
145 | | -1268, 416, -324, -80, 920, 160, 228, 724, 32, -516, 64, |
146 | | 384, 68, -128, 136, 240, 248, -204, -68, 252, -932, -120, |
147 | | -480, -628, -84, 192, 852, -404, -288, -132, 204, 100, 168, |
148 | | -68, -196, -868, 460, 1080, 380, -80, 244, 0, 484, -888, |
149 | | 64, 184, 352, 600, 460, 164, 604, -196, 320, -64, 588, |
150 | | -184, 228, 12, 372, 48, -848, -344, 224, 208, -200, 484, |
151 | | 128, -20, 272, -468, -840, 384, 256, -720, -520, -464, -580, |
152 | | 112, -120, 644, -356, -208, -608, -528, 704, 560, -424, 392, |
153 | | 828, 40, 84, 200, -152, 0, -144, 584, 280, -120, 80, |
154 | | -556, -972, -196, -472, 724, 80, 168, -32, 88, 160, -688, |
155 | | 0, 160, 356, 372, -776, 740, -128, 676, -248, -480, 4, |
156 | | -364, 96, 544, 232, -1032, 956, 236, 356, 20, -40, 300, |
157 | | 24, -676, -596, 132, 1120, -104, 532, -1096, 568, 648, 444, |
158 | | 508, 380, 188, -376, -604, 1488, 424, 24, 756, -220, -192, |
159 | | 716, 120, 920, 688, 168, 44, -460, 568, 284, 1144, 1160, |
160 | | 600, 424, 888, 656, -356, -320, 220, 316, -176, -724, -188, |
161 | | -816, -628, -348, -228, -380, 1012, -452, -660, 736, 928, 404, |
162 | | -696, -72, -268, -892, 128, 184, -344, -780, 360, 336, 400, |
163 | | 344, 428, 548, -112, 136, -228, -216, -820, -516, 340, 92, |
164 | | -136, 116, -300, 376, -244, 100, -316, -520, -284, -12, 824, |
165 | | 164, -548, -180, -128, 116, -924, -828, 268, -368, -580, 620, |
166 | | 192, 160, 0, -1676, 1068, 424, -56, -360, 468, -156, 720, |
167 | | 288, -528, 556, -364, 548, -148, 504, 316, 152, -648, -620, |
168 | | -684, -24, -376, -384, -108, -920, -1032, 768, 180, -264, -508, |
169 | | -1268, -260, -60, 300, -240, 988, 724, -376, -576, -212, -736, |
170 | | 556, 192, 1092, -620, -880, 376, -56, -4, -216, -32, 836, |
171 | | 268, 396, 1332, 864, -600, 100, 56, -412, -92, 356, 180, |
172 | | 884, -468, -436, 292, -388, -804, -704, -840, 368, -348, 140, |
173 | | -724, 1536, 940, 372, 112, -372, 436, -480, 1136, 296, -32, |
174 | | -228, 132, -48, -220, 868, -1016, -60, -1044, -464, 328, 916, |
175 | | 244, 12, -736, -296, 360, 468, -376, -108, -92, 788, 368, |
176 | | -56, 544, 400, -672, -420, 728, 16, 320, 44, -284, -380, |
177 | | -796, 488, 132, 204, -596, -372, 88, -152, -908, -636, -572, |
178 | | -624, -116, -692, -200, -56, 276, -88, 484, -324, 948, 864, |
179 | | 1000, -456, -184, -276, 292, -296, 156, 676, 320, 160, 908, |
180 | | -84, -1236, -288, -116, 260, -372, -644, 732, -756, -96, 84, |
181 | | 344, -520, 348, -688, 240, -84, 216, -1044, -136, -676, -396, |
182 | | -1500, 960, -40, 176, 168, 1516, 420, -504, -344, -364, -360, |
183 | | 1216, -940, -380, -212, 252, -660, -708, 484, -444, -152, 928, |
184 | | -120, 1112, 476, -260, 560, -148, -344, 108, -196, 228, -288, |
185 | | 504, 560, -328, -88, 288, -1008, 460, -228, 468, -836, -196, |
186 | | 76, 388, 232, 412, -1168, -716, -644, 756, -172, -356, -504, |
187 | | 116, 432, 528, 48, 476, -168, -608, 448, 160, -532, -272, |
188 | | 28, -676, -12, 828, 980, 456, 520, 104, -104, 256, -344, |
189 | | -4, -28, -368, -52, -524, -572, -556, -200, 768, 1124, -208, |
190 | | -512, 176, 232, 248, -148, -888, 604, -600, -304, 804, -156, |
191 | | -212, 488, -192, -804, -256, 368, -360, -916, -328, 228, -240, |
192 | | -448, -472, 856, -556, -364, 572, -12, -156, -368, -340, 432, |
193 | | 252, -752, -152, 288, 268, -580, -848, -592, 108, -76, 244, |
194 | | 312, -716, 592, -80, 436, 360, 4, -248, 160, 516, 584, |
195 | | 732, 44, -468, -280, -292, -156, -588, 28, 308, 912, 24, |
196 | | 124, 156, 180, -252, 944, -924, -772, -520, -428, -624, 300, |
197 | | -212, -1144, 32, -724, 800, -1128, -212, -1288, -848, 180, -416, |
198 | | 440, 192, -576, -792, -76, -1080, 80, -532, -352, -132, 380, |
199 | | -820, 148, 1112, 128, 164, 456, 700, -924, 144, -668, -384, |
200 | | 648, -832, 508, 552, -52, -100, -656, 208, -568, 748, -88, |
201 | | 680, 232, 300, 192, -408, -1012, -152, -252, -268, 272, -876, |
202 | | -664, -648, -332, -136, 16, 12, 1152, -28, 332, -536, 320, |
203 | | -672, -460, -316, 532, -260, 228, -40, 1052, -816, 180, 88, |
204 | | -496, -556, -672, -368, 428, 92, 356, 404, -408, 252, 196, |
205 | | -176, -556, 792, 268, 32, 372, 40, 96, -332, 328, 120, |
206 | | 372, -900, -40, 472, -264, -592, 952, 128, 656, 112, 664, |
207 | | -232, 420, 4, -344, -464, 556, 244, -416, -32, 252, 0, |
208 | | -412, 188, -696, 508, -476, 324, -1096, 656, -312, 560, 264, |
209 | | -136, 304, 160, -64, -580, 248, 336, -720, 560, -348, -288, |
210 | | -276, -196, -500, 852, -544, -236, -1128, -992, -776, 116, 56, |
211 | | 52, 860, 884, 212, -12, 168, 1020, 512, -552, 924, -148, |
212 | | 716, 188, 164, -340, -520, -184, 880, -152, -680, -208, -1156, |
213 | | -300, -528, -472, 364, 100, -744, -1056, -32, 540, 280, 144, |
214 | | -676, -32, -232, -280, -224, 96, 568, -76, 172, 148, 148, |
215 | | 104, 32, -296, -32, 788, -80, 32, -16, 280, 288, 944, |
216 | | 428, -484 |
217 | | }; |
218 | | |
219 | | static const int gauss_bits = 11; |
220 | | |
221 | | static const int luma_subblock_size_y = 32; |
222 | | static const int luma_subblock_size_x = 32; |
223 | | |
224 | | static const int min_luma_legal_range = 16; |
225 | | static const int max_luma_legal_range = 235; |
226 | | |
227 | | static const int min_chroma_legal_range = 16; |
228 | | static const int max_chroma_legal_range = 240; |
229 | | |
230 | | typedef struct { |
231 | | int y[256]; |
232 | | int cb[256]; |
233 | | int cr[256]; |
234 | | } aom_grain_scaling_lut_t; |
235 | | |
236 | | typedef struct { |
237 | | uint16_t random_register; // random number generator register |
238 | | } aom_grain_rng_t; |
239 | | |
240 | | static void dealloc_arrays(const aom_film_grain_t *params, int ***pred_pos_luma, |
241 | | int ***pred_pos_chroma, int **luma_grain_block, |
242 | | int **cb_grain_block, int **cr_grain_block, |
243 | | int **y_line_buf, int **cb_line_buf, |
244 | | int **cr_line_buf, int **y_col_buf, int **cb_col_buf, |
245 | 903 | int **cr_col_buf) { |
246 | 903 | int num_pos_luma = 2 * params->ar_coeff_lag * (params->ar_coeff_lag + 1); |
247 | 903 | int num_pos_chroma = num_pos_luma; |
248 | 903 | if (params->num_y_points > 0) ++num_pos_chroma; |
249 | | |
250 | 903 | if (*pred_pos_luma) { |
251 | 6.53k | for (int row = 0; row < num_pos_luma; row++) { |
252 | 5.63k | aom_free((*pred_pos_luma)[row]); |
253 | 5.63k | } |
254 | 903 | aom_free(*pred_pos_luma); |
255 | 903 | *pred_pos_luma = NULL; |
256 | 903 | } |
257 | | |
258 | 903 | if (*pred_pos_chroma) { |
259 | 6.85k | for (int row = 0; row < num_pos_chroma; row++) { |
260 | 5.94k | aom_free((*pred_pos_chroma)[row]); |
261 | 5.94k | } |
262 | 903 | aom_free(*pred_pos_chroma); |
263 | 903 | *pred_pos_chroma = NULL; |
264 | 903 | } |
265 | | |
266 | 903 | aom_free(*y_line_buf); |
267 | 903 | *y_line_buf = NULL; |
268 | | |
269 | 903 | aom_free(*cb_line_buf); |
270 | 903 | *cb_line_buf = NULL; |
271 | | |
272 | 903 | aom_free(*cr_line_buf); |
273 | 903 | *cr_line_buf = NULL; |
274 | | |
275 | 903 | aom_free(*y_col_buf); |
276 | 903 | *y_col_buf = NULL; |
277 | | |
278 | 903 | aom_free(*cb_col_buf); |
279 | 903 | *cb_col_buf = NULL; |
280 | | |
281 | 903 | aom_free(*cr_col_buf); |
282 | 903 | *cr_col_buf = NULL; |
283 | | |
284 | 903 | aom_free(*luma_grain_block); |
285 | 903 | *luma_grain_block = NULL; |
286 | | |
287 | 903 | aom_free(*cb_grain_block); |
288 | 903 | *cb_grain_block = NULL; |
289 | | |
290 | 903 | aom_free(*cr_grain_block); |
291 | 903 | *cr_grain_block = NULL; |
292 | 903 | } |
293 | | |
294 | | static bool init_arrays(const aom_film_grain_t *params, int luma_stride, |
295 | | int chroma_stride, int ***pred_pos_luma_p, |
296 | | int ***pred_pos_chroma_p, int **luma_grain_block, |
297 | | int **cb_grain_block, int **cr_grain_block, |
298 | | int **y_line_buf, int **cb_line_buf, int **cr_line_buf, |
299 | | int **y_col_buf, int **cb_col_buf, int **cr_col_buf, |
300 | | int luma_grain_samples, int chroma_grain_samples, |
301 | 903 | int chroma_subsamp_y, int chroma_subsamp_x) { |
302 | 903 | *pred_pos_luma_p = NULL; |
303 | 903 | *pred_pos_chroma_p = NULL; |
304 | 903 | *luma_grain_block = NULL; |
305 | 903 | *cb_grain_block = NULL; |
306 | 903 | *cr_grain_block = NULL; |
307 | 903 | *y_line_buf = NULL; |
308 | 903 | *cb_line_buf = NULL; |
309 | 903 | *cr_line_buf = NULL; |
310 | 903 | *y_col_buf = NULL; |
311 | 903 | *cb_col_buf = NULL; |
312 | 903 | *cr_col_buf = NULL; |
313 | | |
314 | 903 | const int chroma_subblock_size_y = luma_subblock_size_y >> chroma_subsamp_y; |
315 | | |
316 | 903 | int num_pos_luma = 2 * params->ar_coeff_lag * (params->ar_coeff_lag + 1); |
317 | 903 | int num_pos_chroma = num_pos_luma; |
318 | 903 | if (params->num_y_points > 0) ++num_pos_chroma; |
319 | | |
320 | 903 | int **pred_pos_luma; |
321 | 903 | int **pred_pos_chroma; |
322 | | |
323 | 903 | pred_pos_luma = (int **)aom_calloc(num_pos_luma, sizeof(*pred_pos_luma)); |
324 | 903 | if (!pred_pos_luma) return false; |
325 | | |
326 | 6.53k | for (int row = 0; row < num_pos_luma; row++) { |
327 | 5.63k | pred_pos_luma[row] = (int *)aom_malloc(sizeof(**pred_pos_luma) * 3); |
328 | 5.63k | if (!pred_pos_luma[row]) { |
329 | 0 | dealloc_arrays(params, pred_pos_luma_p, pred_pos_chroma_p, |
330 | 0 | luma_grain_block, cb_grain_block, cr_grain_block, |
331 | 0 | y_line_buf, cb_line_buf, cr_line_buf, y_col_buf, |
332 | 0 | cb_col_buf, cr_col_buf); |
333 | 0 | return false; |
334 | 0 | } |
335 | 5.63k | } |
336 | | |
337 | 903 | pred_pos_chroma = |
338 | 903 | (int **)aom_calloc(num_pos_chroma, sizeof(*pred_pos_chroma)); |
339 | 903 | if (!pred_pos_chroma) { |
340 | 0 | dealloc_arrays(params, pred_pos_luma_p, pred_pos_chroma_p, luma_grain_block, |
341 | 0 | cb_grain_block, cr_grain_block, y_line_buf, cb_line_buf, |
342 | 0 | cr_line_buf, y_col_buf, cb_col_buf, cr_col_buf); |
343 | 0 | return false; |
344 | 0 | } |
345 | | |
346 | 6.85k | for (int row = 0; row < num_pos_chroma; row++) { |
347 | 5.94k | pred_pos_chroma[row] = (int *)aom_malloc(sizeof(**pred_pos_chroma) * 3); |
348 | 5.94k | if (!pred_pos_chroma[row]) { |
349 | 0 | dealloc_arrays(params, pred_pos_luma_p, pred_pos_chroma_p, |
350 | 0 | luma_grain_block, cb_grain_block, cr_grain_block, |
351 | 0 | y_line_buf, cb_line_buf, cr_line_buf, y_col_buf, |
352 | 0 | cb_col_buf, cr_col_buf); |
353 | 0 | return false; |
354 | 0 | } |
355 | 5.94k | } |
356 | | |
357 | 903 | int pos_ar_index = 0; |
358 | | |
359 | 1.80k | for (int row = -params->ar_coeff_lag; row < 0; row++) { |
360 | 5.63k | for (int col = -params->ar_coeff_lag; col < params->ar_coeff_lag + 1; |
361 | 4.73k | col++) { |
362 | 4.73k | pred_pos_luma[pos_ar_index][0] = row; |
363 | 4.73k | pred_pos_luma[pos_ar_index][1] = col; |
364 | 4.73k | pred_pos_luma[pos_ar_index][2] = 0; |
365 | | |
366 | 4.73k | pred_pos_chroma[pos_ar_index][0] = row; |
367 | 4.73k | pred_pos_chroma[pos_ar_index][1] = col; |
368 | 4.73k | pred_pos_chroma[pos_ar_index][2] = 0; |
369 | 4.73k | ++pos_ar_index; |
370 | 4.73k | } |
371 | 906 | } |
372 | | |
373 | 1.80k | for (int col = -params->ar_coeff_lag; col < 0; col++) { |
374 | 906 | pred_pos_luma[pos_ar_index][0] = 0; |
375 | 906 | pred_pos_luma[pos_ar_index][1] = col; |
376 | 906 | pred_pos_luma[pos_ar_index][2] = 0; |
377 | | |
378 | 906 | pred_pos_chroma[pos_ar_index][0] = 0; |
379 | 906 | pred_pos_chroma[pos_ar_index][1] = col; |
380 | 906 | pred_pos_chroma[pos_ar_index][2] = 0; |
381 | | |
382 | 906 | ++pos_ar_index; |
383 | 906 | } |
384 | | |
385 | 903 | if (params->num_y_points > 0) { |
386 | 312 | pred_pos_chroma[pos_ar_index][0] = 0; |
387 | 312 | pred_pos_chroma[pos_ar_index][1] = 0; |
388 | 312 | pred_pos_chroma[pos_ar_index][2] = 1; |
389 | 312 | } |
390 | | |
391 | 903 | *pred_pos_luma_p = pred_pos_luma; |
392 | 903 | *pred_pos_chroma_p = pred_pos_chroma; |
393 | | |
394 | 903 | *y_line_buf = (int *)aom_malloc(sizeof(**y_line_buf) * luma_stride * 2); |
395 | 903 | *cb_line_buf = (int *)aom_malloc(sizeof(**cb_line_buf) * chroma_stride * |
396 | 903 | (2 >> chroma_subsamp_y)); |
397 | 903 | *cr_line_buf = (int *)aom_malloc(sizeof(**cr_line_buf) * chroma_stride * |
398 | 903 | (2 >> chroma_subsamp_y)); |
399 | | |
400 | 903 | *y_col_buf = |
401 | 903 | (int *)aom_malloc(sizeof(**y_col_buf) * (luma_subblock_size_y + 2) * 2); |
402 | 903 | *cb_col_buf = |
403 | 903 | (int *)aom_malloc(sizeof(**cb_col_buf) * |
404 | 903 | (chroma_subblock_size_y + (2 >> chroma_subsamp_y)) * |
405 | 903 | (2 >> chroma_subsamp_x)); |
406 | 903 | *cr_col_buf = |
407 | 903 | (int *)aom_malloc(sizeof(**cr_col_buf) * |
408 | 903 | (chroma_subblock_size_y + (2 >> chroma_subsamp_y)) * |
409 | 903 | (2 >> chroma_subsamp_x)); |
410 | | |
411 | 903 | *luma_grain_block = |
412 | 903 | (int *)aom_malloc(sizeof(**luma_grain_block) * luma_grain_samples); |
413 | 903 | *cb_grain_block = |
414 | 903 | (int *)aom_malloc(sizeof(**cb_grain_block) * chroma_grain_samples); |
415 | 903 | *cr_grain_block = |
416 | 903 | (int *)aom_malloc(sizeof(**cr_grain_block) * chroma_grain_samples); |
417 | 903 | if (!(*pred_pos_luma_p && *pred_pos_chroma_p && *y_line_buf && *cb_line_buf && |
418 | 903 | *cr_line_buf && *y_col_buf && *cb_col_buf && *cr_col_buf && |
419 | 903 | *luma_grain_block && *cb_grain_block && *cr_grain_block)) { |
420 | 0 | dealloc_arrays(params, pred_pos_luma_p, pred_pos_chroma_p, luma_grain_block, |
421 | 0 | cb_grain_block, cr_grain_block, y_line_buf, cb_line_buf, |
422 | 0 | cr_line_buf, y_col_buf, cb_col_buf, cr_col_buf); |
423 | 0 | return false; |
424 | 0 | } |
425 | 903 | return true; |
426 | 903 | } |
427 | | |
428 | | // get a number between 0 and 2^bits - 1 |
429 | 4.78M | static inline int get_random_number(aom_grain_rng_t *rng, int bits) { |
430 | 4.78M | uint16_t bit; |
431 | 4.78M | bit = ((rng->random_register >> 0) ^ (rng->random_register >> 1) ^ |
432 | 4.78M | (rng->random_register >> 3) ^ (rng->random_register >> 12)) & |
433 | 4.78M | 1; |
434 | 4.78M | rng->random_register = (rng->random_register >> 1) | (bit << 15); |
435 | 4.78M | return (rng->random_register >> (16 - bits)) & ((1 << bits) - 1); |
436 | 4.78M | } |
437 | | |
438 | | static void init_random_generator(aom_grain_rng_t *rng, int luma_line, |
439 | 17.8k | uint16_t seed) { |
440 | | // same for the picture |
441 | | |
442 | 17.8k | uint16_t msb = (seed >> 8) & 255; |
443 | 17.8k | uint16_t lsb = seed & 255; |
444 | | |
445 | 17.8k | rng->random_register = (msb << 8) + lsb; |
446 | | |
447 | | // changes for each row |
448 | 17.8k | int luma_num = luma_line >> 5; |
449 | | |
450 | 17.8k | rng->random_register ^= ((luma_num * 37 + 178) & 255) << 8; |
451 | 17.8k | rng->random_register ^= ((luma_num * 173 + 105) & 255); |
452 | 17.8k | } |
453 | | |
454 | | static void generate_luma_grain_block( |
455 | | const aom_film_grain_t *params, aom_grain_rng_t *rng, int **pred_pos_luma, |
456 | | int *luma_grain_block, int luma_block_size_y, int luma_block_size_x, |
457 | | int luma_grain_stride, int left_pad, int top_pad, int right_pad, |
458 | 903 | int bottom_pad) { |
459 | 903 | if (params->num_y_points == 0) { |
460 | 591 | memset(luma_grain_block, 0, |
461 | 591 | sizeof(*luma_grain_block) * luma_block_size_y * luma_grain_stride); |
462 | 591 | return; |
463 | 591 | } |
464 | | |
465 | 312 | int bit_depth = params->bit_depth; |
466 | 312 | int gauss_sec_shift = 12 - bit_depth + params->grain_scale_shift; |
467 | | |
468 | 312 | int num_pos_luma = 2 * params->ar_coeff_lag * (params->ar_coeff_lag + 1); |
469 | 312 | int rounding_offset = (1 << (params->ar_coeff_shift - 1)); |
470 | | |
471 | 312 | const int grain_min = -(1 << (bit_depth - 1)); |
472 | 312 | const int grain_max = (1 << (bit_depth - 1)) - 1; |
473 | | |
474 | 23.0k | for (int i = 0; i < luma_block_size_y; i++) |
475 | 1.89M | for (int j = 0; j < luma_block_size_x; j++) |
476 | 1.86M | luma_grain_block[i * luma_grain_stride + j] = |
477 | 1.86M | (gaussian_sequence[get_random_number(rng, gauss_bits)] + |
478 | 1.86M | ((1 << gauss_sec_shift) >> 1)) >> |
479 | 1.86M | gauss_sec_shift; |
480 | | |
481 | 22.1k | for (int i = top_pad; i < luma_block_size_y - bottom_pad; i++) |
482 | 1.68M | for (int j = left_pad; j < luma_block_size_x - right_pad; j++) { |
483 | 1.65M | int wsum = 0; |
484 | 10.7M | for (int pos = 0; pos < num_pos_luma; pos++) { |
485 | 9.10M | wsum = wsum + params->ar_coeffs_y[pos] * |
486 | 9.10M | luma_grain_block[(i + pred_pos_luma[pos][0]) * |
487 | 9.10M | luma_grain_stride + |
488 | 9.10M | j + pred_pos_luma[pos][1]]; |
489 | 9.10M | } |
490 | 1.65M | luma_grain_block[i * luma_grain_stride + j] = |
491 | 1.65M | clamp(luma_grain_block[i * luma_grain_stride + j] + |
492 | 1.65M | ((wsum + rounding_offset) >> params->ar_coeff_shift), |
493 | 1.65M | grain_min, grain_max); |
494 | 1.65M | } |
495 | 312 | } |
496 | | |
497 | | static bool generate_chroma_grain_blocks( |
498 | | const aom_film_grain_t *params, aom_grain_rng_t *rng, int **pred_pos_chroma, |
499 | | int *luma_grain_block, int *cb_grain_block, int *cr_grain_block, |
500 | | int luma_grain_stride, int chroma_block_size_y, int chroma_block_size_x, |
501 | | int chroma_grain_stride, int left_pad, int top_pad, int right_pad, |
502 | 903 | int bottom_pad, int chroma_subsamp_y, int chroma_subsamp_x) { |
503 | 903 | int bit_depth = params->bit_depth; |
504 | 903 | int gauss_sec_shift = 12 - bit_depth + params->grain_scale_shift; |
505 | | |
506 | 903 | int num_pos_chroma = 2 * params->ar_coeff_lag * (params->ar_coeff_lag + 1); |
507 | 903 | if (params->num_y_points > 0) ++num_pos_chroma; |
508 | 903 | int rounding_offset = (1 << (params->ar_coeff_shift - 1)); |
509 | 903 | int chroma_grain_block_size = chroma_block_size_y * chroma_grain_stride; |
510 | | |
511 | 903 | const int grain_min = -(1 << (bit_depth - 1)); |
512 | 903 | const int grain_max = (1 << (bit_depth - 1)) - 1; |
513 | | |
514 | 903 | if (params->num_cb_points || params->chroma_scaling_from_luma) { |
515 | 371 | init_random_generator(rng, 7 << 5, params->random_seed); |
516 | | |
517 | 22.3k | for (int i = 0; i < chroma_block_size_y; i++) |
518 | 1.52M | for (int j = 0; j < chroma_block_size_x; j++) |
519 | 1.50M | cb_grain_block[i * chroma_grain_stride + j] = |
520 | 1.50M | (gaussian_sequence[get_random_number(rng, gauss_bits)] + |
521 | 1.50M | ((1 << gauss_sec_shift) >> 1)) >> |
522 | 1.50M | gauss_sec_shift; |
523 | 532 | } else { |
524 | 532 | memset(cb_grain_block, 0, |
525 | 532 | sizeof(*cb_grain_block) * chroma_grain_block_size); |
526 | 532 | } |
527 | | |
528 | 903 | if (params->num_cr_points || params->chroma_scaling_from_luma) { |
529 | 344 | init_random_generator(rng, 11 << 5, params->random_seed); |
530 | | |
531 | 20.3k | for (int i = 0; i < chroma_block_size_y; i++) |
532 | 1.35M | for (int j = 0; j < chroma_block_size_x; j++) |
533 | 1.33M | cr_grain_block[i * chroma_grain_stride + j] = |
534 | 1.33M | (gaussian_sequence[get_random_number(rng, gauss_bits)] + |
535 | 1.33M | ((1 << gauss_sec_shift) >> 1)) >> |
536 | 1.33M | gauss_sec_shift; |
537 | 559 | } else { |
538 | 559 | memset(cr_grain_block, 0, |
539 | 559 | sizeof(*cr_grain_block) * chroma_grain_block_size); |
540 | 559 | } |
541 | | |
542 | 43.9k | for (int i = top_pad; i < chroma_block_size_y - bottom_pad; i++) |
543 | 2.40M | for (int j = left_pad; j < chroma_block_size_x - right_pad; j++) { |
544 | 2.36M | int wsum_cb = 0; |
545 | 2.36M | int wsum_cr = 0; |
546 | 21.1M | for (int pos = 0; pos < num_pos_chroma; pos++) { |
547 | 18.7M | if (pred_pos_chroma[pos][2] == 0) { |
548 | 18.0M | wsum_cb = wsum_cb + params->ar_coeffs_cb[pos] * |
549 | 18.0M | cb_grain_block[(i + pred_pos_chroma[pos][0]) * |
550 | 18.0M | chroma_grain_stride + |
551 | 18.0M | j + pred_pos_chroma[pos][1]]; |
552 | 18.0M | wsum_cr = wsum_cr + params->ar_coeffs_cr[pos] * |
553 | 18.0M | cr_grain_block[(i + pred_pos_chroma[pos][0]) * |
554 | 18.0M | chroma_grain_stride + |
555 | 18.0M | j + pred_pos_chroma[pos][1]]; |
556 | 18.0M | } else if (pred_pos_chroma[pos][2] == 1) { |
557 | 684k | int av_luma = 0; |
558 | 684k | int luma_coord_y = ((i - top_pad) << chroma_subsamp_y) + top_pad; |
559 | 684k | int luma_coord_x = ((j - left_pad) << chroma_subsamp_x) + left_pad; |
560 | | |
561 | 1.67M | for (int k = luma_coord_y; k < luma_coord_y + chroma_subsamp_y + 1; |
562 | 994k | k++) |
563 | 2.65M | for (int l = luma_coord_x; l < luma_coord_x + chroma_subsamp_x + 1; |
564 | 1.65M | l++) |
565 | 1.65M | av_luma += luma_grain_block[k * luma_grain_stride + l]; |
566 | | |
567 | 684k | av_luma = |
568 | 684k | (av_luma + ((1 << (chroma_subsamp_y + chroma_subsamp_x)) >> 1)) >> |
569 | 684k | (chroma_subsamp_y + chroma_subsamp_x); |
570 | | |
571 | 684k | wsum_cb = wsum_cb + params->ar_coeffs_cb[pos] * av_luma; |
572 | 684k | wsum_cr = wsum_cr + params->ar_coeffs_cr[pos] * av_luma; |
573 | 684k | } else { |
574 | 0 | fprintf( |
575 | 0 | stderr, |
576 | 0 | "Grain synthesis: prediction between two chroma components is " |
577 | 0 | "not supported!"); |
578 | 0 | return false; |
579 | 0 | } |
580 | 18.7M | } |
581 | 2.36M | if (params->num_cb_points || params->chroma_scaling_from_luma) |
582 | 1.31M | cb_grain_block[i * chroma_grain_stride + j] = |
583 | 1.31M | clamp(cb_grain_block[i * chroma_grain_stride + j] + |
584 | 1.31M | ((wsum_cb + rounding_offset) >> params->ar_coeff_shift), |
585 | 1.31M | grain_min, grain_max); |
586 | 2.36M | if (params->num_cr_points || params->chroma_scaling_from_luma) |
587 | 1.15M | cr_grain_block[i * chroma_grain_stride + j] = |
588 | 1.15M | clamp(cr_grain_block[i * chroma_grain_stride + j] + |
589 | 1.15M | ((wsum_cr + rounding_offset) >> params->ar_coeff_shift), |
590 | 1.15M | grain_min, grain_max); |
591 | 2.36M | } |
592 | 903 | return true; |
593 | 903 | } |
594 | | |
595 | | static void init_scaling_function(const int scaling_points[][2], int num_points, |
596 | 2.21k | int scaling_lut[]) { |
597 | 2.21k | if (num_points == 0) return; |
598 | | |
599 | 39.9k | for (int i = 0; i < scaling_points[0][0]; i++) |
600 | 39.3k | scaling_lut[i] = scaling_points[0][1]; |
601 | | |
602 | 829 | for (int point = 0; point < num_points - 1; point++) { |
603 | 300 | int delta_y = scaling_points[point + 1][1] - scaling_points[point][1]; |
604 | 300 | int delta_x = scaling_points[point + 1][0] - scaling_points[point][0]; |
605 | | |
606 | 300 | int64_t delta = delta_y * ((65536 + (delta_x >> 1)) / delta_x); |
607 | | |
608 | 28.4k | for (int x = 0; x < delta_x; x++) { |
609 | 28.1k | scaling_lut[scaling_points[point][0] + x] = |
610 | 28.1k | scaling_points[point][1] + (int)((x * delta + 32768) >> 16); |
611 | 28.1k | } |
612 | 300 | } |
613 | | |
614 | 68.4k | for (int i = scaling_points[num_points - 1][0]; i < 256; i++) |
615 | 67.9k | scaling_lut[i] = scaling_points[num_points - 1][1]; |
616 | 529 | } |
617 | | |
618 | | // function that extracts samples from a LUT (and interpolates intemediate |
619 | | // frames for 10- and 12-bit video) |
620 | 97.8M | static int scale_LUT(const int *scaling_lut, int index, int bit_depth) { |
621 | 97.8M | int x = index >> (bit_depth - 8); |
622 | | |
623 | 97.8M | if (!(bit_depth - 8) || x == 255) |
624 | 74.9M | return scaling_lut[x]; |
625 | 22.8M | else |
626 | 22.8M | return scaling_lut[x] + (((scaling_lut[x + 1] - scaling_lut[x]) * |
627 | 22.8M | (index & ((1 << (bit_depth - 8)) - 1)) + |
628 | 22.8M | (1 << (bit_depth - 9))) >> |
629 | 22.8M | (bit_depth - 8)); |
630 | 97.8M | } |
631 | | |
632 | | static void add_noise_to_block(const aom_film_grain_t *params, |
633 | | const aom_grain_scaling_lut_t *scaling_lut, |
634 | | uint8_t *luma, uint8_t *cb, uint8_t *cr, |
635 | | int luma_stride, int chroma_stride, |
636 | | int *luma_grain, int *cb_grain, int *cr_grain, |
637 | | int luma_grain_stride, int chroma_grain_stride, |
638 | | int half_luma_height, int half_luma_width, |
639 | | int bit_depth, int chroma_subsamp_y, |
640 | 103k | int chroma_subsamp_x, int mc_identity) { |
641 | 103k | int cb_mult = params->cb_mult - 128; // fixed scale |
642 | 103k | int cb_luma_mult = params->cb_luma_mult - 128; // fixed scale |
643 | 103k | int cb_offset = params->cb_offset - 256; |
644 | | |
645 | 103k | int cr_mult = params->cr_mult - 128; // fixed scale |
646 | 103k | int cr_luma_mult = params->cr_luma_mult - 128; // fixed scale |
647 | 103k | int cr_offset = params->cr_offset - 256; |
648 | | |
649 | 103k | int rounding_offset = (1 << (params->scaling_shift - 1)); |
650 | | |
651 | 103k | int apply_y = params->num_y_points > 0 ? 1 : 0; |
652 | 103k | int apply_cb = |
653 | 103k | (params->num_cb_points > 0 || params->chroma_scaling_from_luma) ? 1 : 0; |
654 | 103k | int apply_cr = |
655 | 103k | (params->num_cr_points > 0 || params->chroma_scaling_from_luma) ? 1 : 0; |
656 | | |
657 | 103k | if (params->chroma_scaling_from_luma) { |
658 | 79.0k | cb_mult = 0; // fixed scale |
659 | 79.0k | cb_luma_mult = 64; // fixed scale |
660 | 79.0k | cb_offset = 0; |
661 | | |
662 | 79.0k | cr_mult = 0; // fixed scale |
663 | 79.0k | cr_luma_mult = 64; // fixed scale |
664 | 79.0k | cr_offset = 0; |
665 | 79.0k | } |
666 | | |
667 | 103k | int min_luma, max_luma, min_chroma, max_chroma; |
668 | | |
669 | 103k | if (params->clip_to_restricted_range) { |
670 | 13.1k | min_luma = min_luma_legal_range; |
671 | 13.1k | max_luma = max_luma_legal_range; |
672 | | |
673 | 13.1k | if (mc_identity) { |
674 | 8 | min_chroma = min_luma_legal_range; |
675 | 8 | max_chroma = max_luma_legal_range; |
676 | 13.1k | } else { |
677 | 13.1k | min_chroma = min_chroma_legal_range; |
678 | 13.1k | max_chroma = max_chroma_legal_range; |
679 | 13.1k | } |
680 | 90.2k | } else { |
681 | 90.2k | min_luma = min_chroma = 0; |
682 | 90.2k | max_luma = max_chroma = 255; |
683 | 90.2k | } |
684 | | |
685 | 1.89M | for (int i = 0; i < (half_luma_height << (1 - chroma_subsamp_y)); i++) { |
686 | 28.2M | for (int j = 0; j < (half_luma_width << (1 - chroma_subsamp_x)); j++) { |
687 | 26.5M | int average_luma = 0; |
688 | 26.5M | if (chroma_subsamp_x) { |
689 | 3.19M | average_luma = (luma[(i << chroma_subsamp_y) * luma_stride + |
690 | 3.19M | (j << chroma_subsamp_x)] + |
691 | 3.19M | luma[(i << chroma_subsamp_y) * luma_stride + |
692 | 3.19M | (j << chroma_subsamp_x) + 1] + |
693 | 3.19M | 1) >> |
694 | 3.19M | 1; |
695 | 23.3M | } else { |
696 | 23.3M | average_luma = luma[(i << chroma_subsamp_y) * luma_stride + j]; |
697 | 23.3M | } |
698 | | |
699 | 26.5M | if (apply_cb) { |
700 | 23.9M | cb[i * chroma_stride + j] = clamp( |
701 | 23.9M | cb[i * chroma_stride + j] + |
702 | 23.9M | ((scale_LUT(scaling_lut->cb, |
703 | 23.9M | clamp(((average_luma * cb_luma_mult + |
704 | 23.9M | cb_mult * cb[i * chroma_stride + j]) >> |
705 | 23.9M | 6) + |
706 | 23.9M | cb_offset, |
707 | 23.9M | 0, (256 << (bit_depth - 8)) - 1), |
708 | 23.9M | 8) * |
709 | 23.9M | cb_grain[i * chroma_grain_stride + j] + |
710 | 23.9M | rounding_offset) >> |
711 | 23.9M | params->scaling_shift), |
712 | 23.9M | min_chroma, max_chroma); |
713 | 23.9M | } |
714 | | |
715 | 26.5M | if (apply_cr) { |
716 | 23.6M | cr[i * chroma_stride + j] = clamp( |
717 | 23.6M | cr[i * chroma_stride + j] + |
718 | 23.6M | ((scale_LUT(scaling_lut->cr, |
719 | 23.6M | clamp(((average_luma * cr_luma_mult + |
720 | 23.6M | cr_mult * cr[i * chroma_stride + j]) >> |
721 | 23.6M | 6) + |
722 | 23.6M | cr_offset, |
723 | 23.6M | 0, (256 << (bit_depth - 8)) - 1), |
724 | 23.6M | 8) * |
725 | 23.6M | cr_grain[i * chroma_grain_stride + j] + |
726 | 23.6M | rounding_offset) >> |
727 | 23.6M | params->scaling_shift), |
728 | 23.6M | min_chroma, max_chroma); |
729 | 23.6M | } |
730 | 26.5M | } |
731 | 1.79M | } |
732 | | |
733 | 103k | if (apply_y) { |
734 | 1.62M | for (int i = 0; i < (half_luma_height << 1); i++) { |
735 | 28.9M | for (int j = 0; j < (half_luma_width << 1); j++) { |
736 | 27.3M | luma[i * luma_stride + j] = clamp( |
737 | 27.3M | luma[i * luma_stride + j] + |
738 | 27.3M | ((scale_LUT(scaling_lut->y, luma[i * luma_stride + j], 8) * |
739 | 27.3M | luma_grain[i * luma_grain_stride + j] + |
740 | 27.3M | rounding_offset) >> |
741 | 27.3M | params->scaling_shift), |
742 | 27.3M | min_luma, max_luma); |
743 | 27.3M | } |
744 | 1.55M | } |
745 | 77.3k | } |
746 | 103k | } |
747 | | |
748 | | static void add_noise_to_block_hbd( |
749 | | const aom_film_grain_t *params, const aom_grain_scaling_lut_t *scaling_lut, |
750 | | uint16_t *luma, uint16_t *cb, uint16_t *cr, int luma_stride, |
751 | | int chroma_stride, int *luma_grain, int *cb_grain, int *cr_grain, |
752 | | int luma_grain_stride, int chroma_grain_stride, int half_luma_height, |
753 | | int half_luma_width, int bit_depth, int chroma_subsamp_y, |
754 | 81.2k | int chroma_subsamp_x, int mc_identity) { |
755 | 81.2k | int cb_mult = params->cb_mult - 128; // fixed scale |
756 | 81.2k | int cb_luma_mult = params->cb_luma_mult - 128; // fixed scale |
757 | | // offset value depends on the bit depth |
758 | 81.2k | int cb_offset = (params->cb_offset << (bit_depth - 8)) - (1 << bit_depth); |
759 | | |
760 | 81.2k | int cr_mult = params->cr_mult - 128; // fixed scale |
761 | 81.2k | int cr_luma_mult = params->cr_luma_mult - 128; // fixed scale |
762 | | // offset value depends on the bit depth |
763 | 81.2k | int cr_offset = (params->cr_offset << (bit_depth - 8)) - (1 << bit_depth); |
764 | | |
765 | 81.2k | int rounding_offset = (1 << (params->scaling_shift - 1)); |
766 | | |
767 | 81.2k | int apply_y = params->num_y_points > 0 ? 1 : 0; |
768 | 81.2k | int apply_cb = |
769 | 81.2k | (params->num_cb_points > 0 || params->chroma_scaling_from_luma) > 0 ? 1 |
770 | 81.2k | : 0; |
771 | 81.2k | int apply_cr = |
772 | 81.2k | (params->num_cr_points > 0 || params->chroma_scaling_from_luma) > 0 ? 1 |
773 | 81.2k | : 0; |
774 | | |
775 | 81.2k | if (params->chroma_scaling_from_luma) { |
776 | 14.8k | cb_mult = 0; // fixed scale |
777 | 14.8k | cb_luma_mult = 64; // fixed scale |
778 | 14.8k | cb_offset = 0; |
779 | | |
780 | 14.8k | cr_mult = 0; // fixed scale |
781 | 14.8k | cr_luma_mult = 64; // fixed scale |
782 | 14.8k | cr_offset = 0; |
783 | 14.8k | } |
784 | | |
785 | 81.2k | int min_luma, max_luma, min_chroma, max_chroma; |
786 | | |
787 | 81.2k | if (params->clip_to_restricted_range) { |
788 | 21.9k | min_luma = min_luma_legal_range << (bit_depth - 8); |
789 | 21.9k | max_luma = max_luma_legal_range << (bit_depth - 8); |
790 | | |
791 | 21.9k | if (mc_identity) { |
792 | 1 | min_chroma = min_luma_legal_range << (bit_depth - 8); |
793 | 1 | max_chroma = max_luma_legal_range << (bit_depth - 8); |
794 | 21.9k | } else { |
795 | 21.9k | min_chroma = min_chroma_legal_range << (bit_depth - 8); |
796 | 21.9k | max_chroma = max_chroma_legal_range << (bit_depth - 8); |
797 | 21.9k | } |
798 | 59.3k | } else { |
799 | 59.3k | min_luma = min_chroma = 0; |
800 | 59.3k | max_luma = max_chroma = (256 << (bit_depth - 8)) - 1; |
801 | 59.3k | } |
802 | | |
803 | 943k | for (int i = 0; i < (half_luma_height << (1 - chroma_subsamp_y)); i++) { |
804 | 9.47M | for (int j = 0; j < (half_luma_width << (1 - chroma_subsamp_x)); j++) { |
805 | 8.60M | int average_luma = 0; |
806 | 8.60M | if (chroma_subsamp_x) { |
807 | 7.21M | average_luma = (luma[(i << chroma_subsamp_y) * luma_stride + |
808 | 7.21M | (j << chroma_subsamp_x)] + |
809 | 7.21M | luma[(i << chroma_subsamp_y) * luma_stride + |
810 | 7.21M | (j << chroma_subsamp_x) + 1] + |
811 | 7.21M | 1) >> |
812 | 7.21M | 1; |
813 | 7.21M | } else { |
814 | 1.39M | average_luma = luma[(i << chroma_subsamp_y) * luma_stride + j]; |
815 | 1.39M | } |
816 | | |
817 | 8.60M | if (apply_cb) { |
818 | 2.27M | cb[i * chroma_stride + j] = clamp( |
819 | 2.27M | cb[i * chroma_stride + j] + |
820 | 2.27M | ((scale_LUT(scaling_lut->cb, |
821 | 2.27M | clamp(((average_luma * cb_luma_mult + |
822 | 2.27M | cb_mult * cb[i * chroma_stride + j]) >> |
823 | 2.27M | 6) + |
824 | 2.27M | cb_offset, |
825 | 2.27M | 0, (256 << (bit_depth - 8)) - 1), |
826 | 2.27M | bit_depth) * |
827 | 2.27M | cb_grain[i * chroma_grain_stride + j] + |
828 | 2.27M | rounding_offset) >> |
829 | 2.27M | params->scaling_shift), |
830 | 2.27M | min_chroma, max_chroma); |
831 | 2.27M | } |
832 | 8.60M | if (apply_cr) { |
833 | 2.09M | cr[i * chroma_stride + j] = clamp( |
834 | 2.09M | cr[i * chroma_stride + j] + |
835 | 2.09M | ((scale_LUT(scaling_lut->cr, |
836 | 2.09M | clamp(((average_luma * cr_luma_mult + |
837 | 2.09M | cr_mult * cr[i * chroma_stride + j]) >> |
838 | 2.09M | 6) + |
839 | 2.09M | cr_offset, |
840 | 2.09M | 0, (256 << (bit_depth - 8)) - 1), |
841 | 2.09M | bit_depth) * |
842 | 2.09M | cr_grain[i * chroma_grain_stride + j] + |
843 | 2.09M | rounding_offset) >> |
844 | 2.09M | params->scaling_shift), |
845 | 2.09M | min_chroma, max_chroma); |
846 | 2.09M | } |
847 | 8.60M | } |
848 | 862k | } |
849 | | |
850 | 81.2k | if (apply_y) { |
851 | 1.17M | for (int i = 0; i < (half_luma_height << 1); i++) { |
852 | 19.6M | for (int j = 0; j < (half_luma_width << 1); j++) { |
853 | 18.4M | luma[i * luma_stride + j] = |
854 | 18.4M | clamp(luma[i * luma_stride + j] + |
855 | 18.4M | ((scale_LUT(scaling_lut->y, luma[i * luma_stride + j], |
856 | 18.4M | bit_depth) * |
857 | 18.4M | luma_grain[i * luma_grain_stride + j] + |
858 | 18.4M | rounding_offset) >> |
859 | 18.4M | params->scaling_shift), |
860 | 18.4M | min_luma, max_luma); |
861 | 18.4M | } |
862 | 1.11M | } |
863 | 56.1k | } |
864 | 81.2k | } |
865 | | |
866 | | static void copy_rect(uint8_t *src, int src_stride, uint8_t *dst, |
867 | | int dst_stride, int width, int height, |
868 | 2.01k | int use_high_bit_depth) { |
869 | 2.01k | int hbd_coeff = use_high_bit_depth ? 2 : 1; |
870 | 1.34M | while (height) { |
871 | 1.34M | memcpy(dst, src, width * sizeof(uint8_t) * hbd_coeff); |
872 | 1.34M | src += src_stride; |
873 | 1.34M | dst += dst_stride; |
874 | 1.34M | --height; |
875 | 1.34M | } |
876 | 2.01k | return; |
877 | 2.01k | } |
878 | | |
879 | | static void copy_area(int *src, int src_stride, int *dst, int dst_stride, |
880 | 511k | int width, int height) { |
881 | 5.98M | while (height) { |
882 | 5.47M | memcpy(dst, src, width * sizeof(*src)); |
883 | 5.47M | src += src_stride; |
884 | 5.47M | dst += dst_stride; |
885 | 5.47M | --height; |
886 | 5.47M | } |
887 | 511k | return; |
888 | 511k | } |
889 | | |
890 | | static void extend_even(uint8_t *dst, int dst_stride, int width, int height, |
891 | 903 | int use_high_bit_depth) { |
892 | 903 | if ((width & 1) == 0 && (height & 1) == 0) return; |
893 | 784 | if (use_high_bit_depth) { |
894 | 479 | uint16_t *dst16 = (uint16_t *)dst; |
895 | 479 | int dst16_stride = dst_stride / 2; |
896 | 479 | if (width & 1) { |
897 | 107k | for (int i = 0; i < height; ++i) |
898 | 106k | dst16[i * dst16_stride + width] = dst16[i * dst16_stride + width - 1]; |
899 | 385 | } |
900 | 479 | width = (width + 1) & (~1); |
901 | 479 | if (height & 1) { |
902 | 199 | memcpy(&dst16[height * dst16_stride], &dst16[(height - 1) * dst16_stride], |
903 | 199 | sizeof(*dst16) * width); |
904 | 199 | } |
905 | 479 | } else { |
906 | 305 | if (width & 1) { |
907 | 175k | for (int i = 0; i < height; ++i) |
908 | 174k | dst[i * dst_stride + width] = dst[i * dst_stride + width - 1]; |
909 | 224 | } |
910 | 305 | width = (width + 1) & (~1); |
911 | 305 | if (height & 1) { |
912 | 178 | memcpy(&dst[height * dst_stride], &dst[(height - 1) * dst_stride], |
913 | 178 | sizeof(*dst) * width); |
914 | 178 | } |
915 | 305 | } |
916 | 784 | } |
917 | | |
918 | | static void ver_boundary_overlap(int *left_block, int left_stride, |
919 | | int *right_block, int right_stride, |
920 | | int *dst_block, int dst_stride, int width, |
921 | 142k | int height, int grain_min, int grain_max) { |
922 | 142k | if (width == 1) { |
923 | 899k | while (height) { |
924 | 844k | *dst_block = clamp((*left_block * 23 + *right_block * 22 + 16) >> 5, |
925 | 844k | grain_min, grain_max); |
926 | 844k | left_block += left_stride; |
927 | 844k | right_block += right_stride; |
928 | 844k | dst_block += dst_stride; |
929 | 844k | --height; |
930 | 844k | } |
931 | 54.8k | return; |
932 | 87.4k | } else if (width == 2) { |
933 | 2.89M | while (height) { |
934 | 2.80M | dst_block[0] = clamp((27 * left_block[0] + 17 * right_block[0] + 16) >> 5, |
935 | 2.80M | grain_min, grain_max); |
936 | 2.80M | dst_block[1] = clamp((17 * left_block[1] + 27 * right_block[1] + 16) >> 5, |
937 | 2.80M | grain_min, grain_max); |
938 | 2.80M | left_block += left_stride; |
939 | 2.80M | right_block += right_stride; |
940 | 2.80M | dst_block += dst_stride; |
941 | 2.80M | --height; |
942 | 2.80M | } |
943 | 87.4k | return; |
944 | 87.4k | } |
945 | 142k | } |
946 | | |
947 | | static void hor_boundary_overlap(int *top_block, int top_stride, |
948 | | int *bottom_block, int bottom_stride, |
949 | | int *dst_block, int dst_stride, int width, |
950 | 292k | int height, int grain_min, int grain_max) { |
951 | 292k | if (height == 1) { |
952 | 878k | while (width) { |
953 | 780k | *dst_block = clamp((*top_block * 23 + *bottom_block * 22 + 16) >> 5, |
954 | 780k | grain_min, grain_max); |
955 | 780k | ++top_block; |
956 | 780k | ++bottom_block; |
957 | 780k | ++dst_block; |
958 | 780k | --width; |
959 | 780k | } |
960 | 98.0k | return; |
961 | 194k | } else if (height == 2) { |
962 | 3.22M | while (width) { |
963 | 3.02M | dst_block[0] = clamp((27 * top_block[0] + 17 * bottom_block[0] + 16) >> 5, |
964 | 3.02M | grain_min, grain_max); |
965 | 3.02M | dst_block[dst_stride] = clamp((17 * top_block[top_stride] + |
966 | 3.02M | 27 * bottom_block[bottom_stride] + 16) >> |
967 | 3.02M | 5, |
968 | 3.02M | grain_min, grain_max); |
969 | 3.02M | ++top_block; |
970 | 3.02M | ++bottom_block; |
971 | 3.02M | ++dst_block; |
972 | 3.02M | --width; |
973 | 3.02M | } |
974 | 194k | return; |
975 | 194k | } |
976 | 292k | } |
977 | | |
978 | | /*!\brief Add film grain |
979 | | * |
980 | | * Add film grain to an image |
981 | | * |
982 | | * Returns 0 for success, -1 for failure |
983 | | * |
984 | | * \param[in] grain_params Grain parameters |
985 | | * \param[in] luma luma plane |
986 | | * \param[in] cb cb plane |
987 | | * \param[in] cr cr plane |
988 | | * \param[in] height luma plane height |
989 | | * \param[in] width luma plane width |
990 | | * \param[in] luma_stride luma plane stride |
991 | | * \param[in] chroma_stride chroma plane stride |
992 | | */ |
993 | | static int add_film_grain_run(const aom_film_grain_t *params, uint8_t *luma, |
994 | | uint8_t *cb, uint8_t *cr, int height, int width, |
995 | | int luma_stride, int chroma_stride, |
996 | | int use_high_bit_depth, int chroma_subsamp_y, |
997 | 903 | int chroma_subsamp_x, int mc_identity) { |
998 | 903 | int **pred_pos_luma; |
999 | 903 | int **pred_pos_chroma; |
1000 | 903 | int *luma_grain_block; |
1001 | 903 | int *cb_grain_block; |
1002 | 903 | int *cr_grain_block; |
1003 | | |
1004 | 903 | int *y_line_buf; |
1005 | 903 | int *cb_line_buf; |
1006 | 903 | int *cr_line_buf; |
1007 | | |
1008 | 903 | int *y_col_buf; |
1009 | 903 | int *cb_col_buf; |
1010 | 903 | int *cr_col_buf; |
1011 | | |
1012 | 903 | aom_grain_scaling_lut_t scaling_lut; |
1013 | 903 | memset(&scaling_lut, 0, sizeof(scaling_lut)); |
1014 | | |
1015 | 903 | aom_grain_rng_t rng; |
1016 | 903 | rng.random_register = params->random_seed; |
1017 | | |
1018 | 903 | int left_pad = 3; |
1019 | 903 | int right_pad = 3; // padding to offset for AR coefficients |
1020 | 903 | int top_pad = 3; |
1021 | 903 | int bottom_pad = 0; |
1022 | | |
1023 | 903 | int ar_padding = 3; // maximum lag used for stabilization of AR coefficients |
1024 | | |
1025 | 903 | const int chroma_subblock_size_y = luma_subblock_size_y >> chroma_subsamp_y; |
1026 | 903 | const int chroma_subblock_size_x = luma_subblock_size_x >> chroma_subsamp_x; |
1027 | | |
1028 | | // Initial padding is only needed for generation of |
1029 | | // film grain templates (to stabilize the AR process) |
1030 | | // Only a 64x64 luma and 32x32 chroma part of a template |
1031 | | // is used later for adding grain, padding can be discarded |
1032 | | |
1033 | 903 | int luma_block_size_y = |
1034 | 903 | top_pad + 2 * ar_padding + luma_subblock_size_y * 2 + bottom_pad; |
1035 | 903 | int luma_block_size_x = left_pad + 2 * ar_padding + luma_subblock_size_x * 2 + |
1036 | 903 | 2 * ar_padding + right_pad; |
1037 | | |
1038 | 903 | int chroma_block_size_y = top_pad + (2 >> chroma_subsamp_y) * ar_padding + |
1039 | 903 | chroma_subblock_size_y * 2 + bottom_pad; |
1040 | 903 | int chroma_block_size_x = left_pad + (2 >> chroma_subsamp_x) * ar_padding + |
1041 | 903 | chroma_subblock_size_x * 2 + |
1042 | 903 | (2 >> chroma_subsamp_x) * ar_padding + right_pad; |
1043 | | |
1044 | 903 | int luma_grain_stride = luma_block_size_x; |
1045 | 903 | int chroma_grain_stride = chroma_block_size_x; |
1046 | | |
1047 | 903 | int overlap = params->overlap_flag; |
1048 | 903 | int bit_depth = params->bit_depth; |
1049 | | |
1050 | 903 | int grain_min = -(1 << (bit_depth - 1)); |
1051 | 903 | int grain_max = (1 << (bit_depth - 1)) - 1; |
1052 | | |
1053 | 903 | if (!init_arrays(params, luma_stride, chroma_stride, &pred_pos_luma, |
1054 | 903 | &pred_pos_chroma, &luma_grain_block, &cb_grain_block, |
1055 | 903 | &cr_grain_block, &y_line_buf, &cb_line_buf, &cr_line_buf, |
1056 | 903 | &y_col_buf, &cb_col_buf, &cr_col_buf, |
1057 | 903 | luma_block_size_y * luma_block_size_x, |
1058 | 903 | chroma_block_size_y * chroma_block_size_x, chroma_subsamp_y, |
1059 | 903 | chroma_subsamp_x)) |
1060 | 0 | return -1; |
1061 | | |
1062 | 903 | generate_luma_grain_block(params, &rng, pred_pos_luma, luma_grain_block, |
1063 | 903 | luma_block_size_y, luma_block_size_x, |
1064 | 903 | luma_grain_stride, left_pad, top_pad, right_pad, |
1065 | 903 | bottom_pad); |
1066 | | |
1067 | 903 | if (!generate_chroma_grain_blocks( |
1068 | 903 | params, &rng, pred_pos_chroma, luma_grain_block, cb_grain_block, |
1069 | 903 | cr_grain_block, luma_grain_stride, chroma_block_size_y, |
1070 | 903 | chroma_block_size_x, chroma_grain_stride, left_pad, top_pad, |
1071 | 903 | right_pad, bottom_pad, chroma_subsamp_y, chroma_subsamp_x)) |
1072 | 0 | return -1; |
1073 | | |
1074 | 903 | init_scaling_function(params->scaling_points_y, params->num_y_points, |
1075 | 903 | scaling_lut.y); |
1076 | | |
1077 | 903 | if (params->chroma_scaling_from_luma) { |
1078 | 249 | static_assert(sizeof(scaling_lut.cb) == sizeof(scaling_lut.y), ""); |
1079 | 249 | static_assert(sizeof(scaling_lut.cr) == sizeof(scaling_lut.y), ""); |
1080 | 249 | memcpy(scaling_lut.cb, scaling_lut.y, sizeof(scaling_lut.y)); |
1081 | 249 | memcpy(scaling_lut.cr, scaling_lut.y, sizeof(scaling_lut.y)); |
1082 | 654 | } else { |
1083 | 654 | init_scaling_function(params->scaling_points_cb, params->num_cb_points, |
1084 | 654 | scaling_lut.cb); |
1085 | 654 | init_scaling_function(params->scaling_points_cr, params->num_cr_points, |
1086 | 654 | scaling_lut.cr); |
1087 | 654 | } |
1088 | 18.0k | for (int y = 0; y < height / 2; y += (luma_subblock_size_y >> 1)) { |
1089 | 17.1k | init_random_generator(&rng, y * 2, params->random_seed); |
1090 | | |
1091 | 98.7k | for (int x = 0; x < width / 2; x += (luma_subblock_size_x >> 1)) { |
1092 | 81.6k | int offset_y = get_random_number(&rng, 8); |
1093 | 81.6k | int offset_x = (offset_y >> 4) & 15; |
1094 | 81.6k | offset_y &= 15; |
1095 | | |
1096 | 81.6k | int luma_offset_y = left_pad + 2 * ar_padding + (offset_y << 1); |
1097 | 81.6k | int luma_offset_x = top_pad + 2 * ar_padding + (offset_x << 1); |
1098 | | |
1099 | 81.6k | int chroma_offset_y = top_pad + (2 >> chroma_subsamp_y) * ar_padding + |
1100 | 81.6k | offset_y * (2 >> chroma_subsamp_y); |
1101 | 81.6k | int chroma_offset_x = left_pad + (2 >> chroma_subsamp_x) * ar_padding + |
1102 | 81.6k | offset_x * (2 >> chroma_subsamp_x); |
1103 | | |
1104 | 81.6k | if (overlap && x) { |
1105 | 47.4k | ver_boundary_overlap( |
1106 | 47.4k | y_col_buf, 2, |
1107 | 47.4k | luma_grain_block + luma_offset_y * luma_grain_stride + |
1108 | 47.4k | luma_offset_x, |
1109 | 47.4k | luma_grain_stride, y_col_buf, 2, 2, |
1110 | 47.4k | AOMMIN(luma_subblock_size_y + 2, height - (y << 1)), grain_min, |
1111 | 47.4k | grain_max); |
1112 | | |
1113 | 47.4k | ver_boundary_overlap( |
1114 | 47.4k | cb_col_buf, 2 >> chroma_subsamp_x, |
1115 | 47.4k | cb_grain_block + chroma_offset_y * chroma_grain_stride + |
1116 | 47.4k | chroma_offset_x, |
1117 | 47.4k | chroma_grain_stride, cb_col_buf, 2 >> chroma_subsamp_x, |
1118 | 47.4k | 2 >> chroma_subsamp_x, |
1119 | 47.4k | AOMMIN(chroma_subblock_size_y + (2 >> chroma_subsamp_y), |
1120 | 47.4k | (height - (y << 1)) >> chroma_subsamp_y), |
1121 | 47.4k | grain_min, grain_max); |
1122 | | |
1123 | 47.4k | ver_boundary_overlap( |
1124 | 47.4k | cr_col_buf, 2 >> chroma_subsamp_x, |
1125 | 47.4k | cr_grain_block + chroma_offset_y * chroma_grain_stride + |
1126 | 47.4k | chroma_offset_x, |
1127 | 47.4k | chroma_grain_stride, cr_col_buf, 2 >> chroma_subsamp_x, |
1128 | 47.4k | 2 >> chroma_subsamp_x, |
1129 | 47.4k | AOMMIN(chroma_subblock_size_y + (2 >> chroma_subsamp_y), |
1130 | 47.4k | (height - (y << 1)) >> chroma_subsamp_y), |
1131 | 47.4k | grain_min, grain_max); |
1132 | | |
1133 | 47.4k | int i = y ? 1 : 0; |
1134 | | |
1135 | 47.4k | if (use_high_bit_depth) { |
1136 | 22.8k | add_noise_to_block_hbd( |
1137 | 22.8k | params, &scaling_lut, |
1138 | 22.8k | (uint16_t *)luma + ((y + i) << 1) * luma_stride + (x << 1), |
1139 | 22.8k | (uint16_t *)cb + |
1140 | 22.8k | ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1141 | 22.8k | (x << (1 - chroma_subsamp_x)), |
1142 | 22.8k | (uint16_t *)cr + |
1143 | 22.8k | ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1144 | 22.8k | (x << (1 - chroma_subsamp_x)), |
1145 | 22.8k | luma_stride, chroma_stride, y_col_buf + i * 4, |
1146 | 22.8k | cb_col_buf + i * (2 - chroma_subsamp_y) * (2 - chroma_subsamp_x), |
1147 | 22.8k | cr_col_buf + i * (2 - chroma_subsamp_y) * (2 - chroma_subsamp_x), |
1148 | 22.8k | 2, (2 - chroma_subsamp_x), |
1149 | 22.8k | AOMMIN(luma_subblock_size_y >> 1, height / 2 - y) - i, 1, |
1150 | 22.8k | bit_depth, chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1151 | 24.6k | } else { |
1152 | 24.6k | add_noise_to_block( |
1153 | 24.6k | params, &scaling_lut, |
1154 | 24.6k | luma + ((y + i) << 1) * luma_stride + (x << 1), |
1155 | 24.6k | cb + ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1156 | 24.6k | (x << (1 - chroma_subsamp_x)), |
1157 | 24.6k | cr + ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1158 | 24.6k | (x << (1 - chroma_subsamp_x)), |
1159 | 24.6k | luma_stride, chroma_stride, y_col_buf + i * 4, |
1160 | 24.6k | cb_col_buf + i * (2 - chroma_subsamp_y) * (2 - chroma_subsamp_x), |
1161 | 24.6k | cr_col_buf + i * (2 - chroma_subsamp_y) * (2 - chroma_subsamp_x), |
1162 | 24.6k | 2, (2 - chroma_subsamp_x), |
1163 | 24.6k | AOMMIN(luma_subblock_size_y >> 1, height / 2 - y) - i, 1, |
1164 | 24.6k | bit_depth, chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1165 | 24.6k | } |
1166 | 47.4k | } |
1167 | | |
1168 | 81.6k | if (overlap && y) { |
1169 | 55.5k | if (x) { |
1170 | 42.0k | hor_boundary_overlap(y_line_buf + (x << 1), luma_stride, y_col_buf, 2, |
1171 | 42.0k | y_line_buf + (x << 1), luma_stride, 2, 2, |
1172 | 42.0k | grain_min, grain_max); |
1173 | | |
1174 | 42.0k | hor_boundary_overlap(cb_line_buf + x * (2 >> chroma_subsamp_x), |
1175 | 42.0k | chroma_stride, cb_col_buf, 2 >> chroma_subsamp_x, |
1176 | 42.0k | cb_line_buf + x * (2 >> chroma_subsamp_x), |
1177 | 42.0k | chroma_stride, 2 >> chroma_subsamp_x, |
1178 | 42.0k | 2 >> chroma_subsamp_y, grain_min, grain_max); |
1179 | | |
1180 | 42.0k | hor_boundary_overlap(cr_line_buf + x * (2 >> chroma_subsamp_x), |
1181 | 42.0k | chroma_stride, cr_col_buf, 2 >> chroma_subsamp_x, |
1182 | 42.0k | cr_line_buf + x * (2 >> chroma_subsamp_x), |
1183 | 42.0k | chroma_stride, 2 >> chroma_subsamp_x, |
1184 | 42.0k | 2 >> chroma_subsamp_y, grain_min, grain_max); |
1185 | 42.0k | } |
1186 | | |
1187 | 55.5k | hor_boundary_overlap( |
1188 | 55.5k | y_line_buf + ((x ? x + 1 : 0) << 1), luma_stride, |
1189 | 55.5k | luma_grain_block + luma_offset_y * luma_grain_stride + |
1190 | 55.5k | luma_offset_x + (x ? 2 : 0), |
1191 | 55.5k | luma_grain_stride, y_line_buf + ((x ? x + 1 : 0) << 1), luma_stride, |
1192 | 55.5k | AOMMIN(luma_subblock_size_x - ((x ? 1 : 0) << 1), |
1193 | 55.5k | width - ((x ? x + 1 : 0) << 1)), |
1194 | 55.5k | 2, grain_min, grain_max); |
1195 | | |
1196 | 55.5k | hor_boundary_overlap( |
1197 | 55.5k | cb_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1198 | 55.5k | chroma_stride, |
1199 | 55.5k | cb_grain_block + chroma_offset_y * chroma_grain_stride + |
1200 | 55.5k | chroma_offset_x + ((x ? 1 : 0) << (1 - chroma_subsamp_x)), |
1201 | 55.5k | chroma_grain_stride, |
1202 | 55.5k | cb_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1203 | 55.5k | chroma_stride, |
1204 | 55.5k | AOMMIN(chroma_subblock_size_x - |
1205 | 55.5k | ((x ? 1 : 0) << (1 - chroma_subsamp_x)), |
1206 | 55.5k | (width - ((x ? x + 1 : 0) << 1)) >> chroma_subsamp_x), |
1207 | 55.5k | 2 >> chroma_subsamp_y, grain_min, grain_max); |
1208 | | |
1209 | 55.5k | hor_boundary_overlap( |
1210 | 55.5k | cr_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1211 | 55.5k | chroma_stride, |
1212 | 55.5k | cr_grain_block + chroma_offset_y * chroma_grain_stride + |
1213 | 55.5k | chroma_offset_x + ((x ? 1 : 0) << (1 - chroma_subsamp_x)), |
1214 | 55.5k | chroma_grain_stride, |
1215 | 55.5k | cr_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1216 | 55.5k | chroma_stride, |
1217 | 55.5k | AOMMIN(chroma_subblock_size_x - |
1218 | 55.5k | ((x ? 1 : 0) << (1 - chroma_subsamp_x)), |
1219 | 55.5k | (width - ((x ? x + 1 : 0) << 1)) >> chroma_subsamp_x), |
1220 | 55.5k | 2 >> chroma_subsamp_y, grain_min, grain_max); |
1221 | | |
1222 | 55.5k | if (use_high_bit_depth) { |
1223 | 22.3k | add_noise_to_block_hbd( |
1224 | 22.3k | params, &scaling_lut, |
1225 | 22.3k | (uint16_t *)luma + (y << 1) * luma_stride + (x << 1), |
1226 | 22.3k | (uint16_t *)cb + (y << (1 - chroma_subsamp_y)) * chroma_stride + |
1227 | 22.3k | (x << ((1 - chroma_subsamp_x))), |
1228 | 22.3k | (uint16_t *)cr + (y << (1 - chroma_subsamp_y)) * chroma_stride + |
1229 | 22.3k | (x << ((1 - chroma_subsamp_x))), |
1230 | 22.3k | luma_stride, chroma_stride, y_line_buf + (x << 1), |
1231 | 22.3k | cb_line_buf + (x << (1 - chroma_subsamp_x)), |
1232 | 22.3k | cr_line_buf + (x << (1 - chroma_subsamp_x)), luma_stride, |
1233 | 22.3k | chroma_stride, 1, |
1234 | 22.3k | AOMMIN(luma_subblock_size_x >> 1, width / 2 - x), bit_depth, |
1235 | 22.3k | chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1236 | 33.1k | } else { |
1237 | 33.1k | add_noise_to_block( |
1238 | 33.1k | params, &scaling_lut, luma + (y << 1) * luma_stride + (x << 1), |
1239 | 33.1k | cb + (y << (1 - chroma_subsamp_y)) * chroma_stride + |
1240 | 33.1k | (x << ((1 - chroma_subsamp_x))), |
1241 | 33.1k | cr + (y << (1 - chroma_subsamp_y)) * chroma_stride + |
1242 | 33.1k | (x << ((1 - chroma_subsamp_x))), |
1243 | 33.1k | luma_stride, chroma_stride, y_line_buf + (x << 1), |
1244 | 33.1k | cb_line_buf + (x << (1 - chroma_subsamp_x)), |
1245 | 33.1k | cr_line_buf + (x << (1 - chroma_subsamp_x)), luma_stride, |
1246 | 33.1k | chroma_stride, 1, |
1247 | 33.1k | AOMMIN(luma_subblock_size_x >> 1, width / 2 - x), bit_depth, |
1248 | 33.1k | chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1249 | 33.1k | } |
1250 | 55.5k | } |
1251 | | |
1252 | 81.6k | int i = overlap && y ? 1 : 0; |
1253 | 81.6k | int j = overlap && x ? 1 : 0; |
1254 | | |
1255 | 81.6k | if (use_high_bit_depth) { |
1256 | 36.0k | add_noise_to_block_hbd( |
1257 | 36.0k | params, &scaling_lut, |
1258 | 36.0k | (uint16_t *)luma + ((y + i) << 1) * luma_stride + ((x + j) << 1), |
1259 | 36.0k | (uint16_t *)cb + |
1260 | 36.0k | ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1261 | 36.0k | ((x + j) << (1 - chroma_subsamp_x)), |
1262 | 36.0k | (uint16_t *)cr + |
1263 | 36.0k | ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1264 | 36.0k | ((x + j) << (1 - chroma_subsamp_x)), |
1265 | 36.0k | luma_stride, chroma_stride, |
1266 | 36.0k | luma_grain_block + (luma_offset_y + (i << 1)) * luma_grain_stride + |
1267 | 36.0k | luma_offset_x + (j << 1), |
1268 | 36.0k | cb_grain_block + |
1269 | 36.0k | (chroma_offset_y + (i << (1 - chroma_subsamp_y))) * |
1270 | 36.0k | chroma_grain_stride + |
1271 | 36.0k | chroma_offset_x + (j << (1 - chroma_subsamp_x)), |
1272 | 36.0k | cr_grain_block + |
1273 | 36.0k | (chroma_offset_y + (i << (1 - chroma_subsamp_y))) * |
1274 | 36.0k | chroma_grain_stride + |
1275 | 36.0k | chroma_offset_x + (j << (1 - chroma_subsamp_x)), |
1276 | 36.0k | luma_grain_stride, chroma_grain_stride, |
1277 | 36.0k | AOMMIN(luma_subblock_size_y >> 1, height / 2 - y) - i, |
1278 | 36.0k | AOMMIN(luma_subblock_size_x >> 1, width / 2 - x) - j, bit_depth, |
1279 | 36.0k | chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1280 | 45.5k | } else { |
1281 | 45.5k | add_noise_to_block( |
1282 | 45.5k | params, &scaling_lut, |
1283 | 45.5k | luma + ((y + i) << 1) * luma_stride + ((x + j) << 1), |
1284 | 45.5k | cb + ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1285 | 45.5k | ((x + j) << (1 - chroma_subsamp_x)), |
1286 | 45.5k | cr + ((y + i) << (1 - chroma_subsamp_y)) * chroma_stride + |
1287 | 45.5k | ((x + j) << (1 - chroma_subsamp_x)), |
1288 | 45.5k | luma_stride, chroma_stride, |
1289 | 45.5k | luma_grain_block + (luma_offset_y + (i << 1)) * luma_grain_stride + |
1290 | 45.5k | luma_offset_x + (j << 1), |
1291 | 45.5k | cb_grain_block + |
1292 | 45.5k | (chroma_offset_y + (i << (1 - chroma_subsamp_y))) * |
1293 | 45.5k | chroma_grain_stride + |
1294 | 45.5k | chroma_offset_x + (j << (1 - chroma_subsamp_x)), |
1295 | 45.5k | cr_grain_block + |
1296 | 45.5k | (chroma_offset_y + (i << (1 - chroma_subsamp_y))) * |
1297 | 45.5k | chroma_grain_stride + |
1298 | 45.5k | chroma_offset_x + (j << (1 - chroma_subsamp_x)), |
1299 | 45.5k | luma_grain_stride, chroma_grain_stride, |
1300 | 45.5k | AOMMIN(luma_subblock_size_y >> 1, height / 2 - y) - i, |
1301 | 45.5k | AOMMIN(luma_subblock_size_x >> 1, width / 2 - x) - j, bit_depth, |
1302 | 45.5k | chroma_subsamp_y, chroma_subsamp_x, mc_identity); |
1303 | 45.5k | } |
1304 | | |
1305 | 81.6k | if (overlap) { |
1306 | 61.4k | if (x) { |
1307 | | // Copy overlapped column bufer to line buffer |
1308 | 47.4k | copy_area(y_col_buf + (luma_subblock_size_y << 1), 2, |
1309 | 47.4k | y_line_buf + (x << 1), luma_stride, 2, 2); |
1310 | | |
1311 | 47.4k | copy_area( |
1312 | 47.4k | cb_col_buf + (chroma_subblock_size_y << (1 - chroma_subsamp_x)), |
1313 | 47.4k | 2 >> chroma_subsamp_x, |
1314 | 47.4k | cb_line_buf + (x << (1 - chroma_subsamp_x)), chroma_stride, |
1315 | 47.4k | 2 >> chroma_subsamp_x, 2 >> chroma_subsamp_y); |
1316 | | |
1317 | 47.4k | copy_area( |
1318 | 47.4k | cr_col_buf + (chroma_subblock_size_y << (1 - chroma_subsamp_x)), |
1319 | 47.4k | 2 >> chroma_subsamp_x, |
1320 | 47.4k | cr_line_buf + (x << (1 - chroma_subsamp_x)), chroma_stride, |
1321 | 47.4k | 2 >> chroma_subsamp_x, 2 >> chroma_subsamp_y); |
1322 | 47.4k | } |
1323 | | |
1324 | | // Copy grain to the line buffer for overlap with a bottom block |
1325 | 61.4k | copy_area( |
1326 | 61.4k | luma_grain_block + |
1327 | 61.4k | (luma_offset_y + luma_subblock_size_y) * luma_grain_stride + |
1328 | 61.4k | luma_offset_x + ((x ? 2 : 0)), |
1329 | 61.4k | luma_grain_stride, y_line_buf + ((x ? x + 1 : 0) << 1), luma_stride, |
1330 | 61.4k | AOMMIN(luma_subblock_size_x, width - (x << 1)) - (x ? 2 : 0), 2); |
1331 | | |
1332 | 61.4k | copy_area(cb_grain_block + |
1333 | 61.4k | (chroma_offset_y + chroma_subblock_size_y) * |
1334 | 61.4k | chroma_grain_stride + |
1335 | 61.4k | chroma_offset_x + (x ? 2 >> chroma_subsamp_x : 0), |
1336 | 61.4k | chroma_grain_stride, |
1337 | 61.4k | cb_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1338 | 61.4k | chroma_stride, |
1339 | 61.4k | AOMMIN(chroma_subblock_size_x, |
1340 | 61.4k | ((width - (x << 1)) >> chroma_subsamp_x)) - |
1341 | 61.4k | (x ? 2 >> chroma_subsamp_x : 0), |
1342 | 61.4k | 2 >> chroma_subsamp_y); |
1343 | | |
1344 | 61.4k | copy_area(cr_grain_block + |
1345 | 61.4k | (chroma_offset_y + chroma_subblock_size_y) * |
1346 | 61.4k | chroma_grain_stride + |
1347 | 61.4k | chroma_offset_x + (x ? 2 >> chroma_subsamp_x : 0), |
1348 | 61.4k | chroma_grain_stride, |
1349 | 61.4k | cr_line_buf + ((x ? x + 1 : 0) << (1 - chroma_subsamp_x)), |
1350 | 61.4k | chroma_stride, |
1351 | 61.4k | AOMMIN(chroma_subblock_size_x, |
1352 | 61.4k | ((width - (x << 1)) >> chroma_subsamp_x)) - |
1353 | 61.4k | (x ? 2 >> chroma_subsamp_x : 0), |
1354 | 61.4k | 2 >> chroma_subsamp_y); |
1355 | | |
1356 | | // Copy grain to the column buffer for overlap with the next block to |
1357 | | // the right |
1358 | | |
1359 | 61.4k | copy_area(luma_grain_block + luma_offset_y * luma_grain_stride + |
1360 | 61.4k | luma_offset_x + luma_subblock_size_x, |
1361 | 61.4k | luma_grain_stride, y_col_buf, 2, 2, |
1362 | 61.4k | AOMMIN(luma_subblock_size_y + 2, height - (y << 1))); |
1363 | | |
1364 | 61.4k | copy_area(cb_grain_block + chroma_offset_y * chroma_grain_stride + |
1365 | 61.4k | chroma_offset_x + chroma_subblock_size_x, |
1366 | 61.4k | chroma_grain_stride, cb_col_buf, 2 >> chroma_subsamp_x, |
1367 | 61.4k | 2 >> chroma_subsamp_x, |
1368 | 61.4k | AOMMIN(chroma_subblock_size_y + (2 >> chroma_subsamp_y), |
1369 | 61.4k | (height - (y << 1)) >> chroma_subsamp_y)); |
1370 | | |
1371 | 61.4k | copy_area(cr_grain_block + chroma_offset_y * chroma_grain_stride + |
1372 | 61.4k | chroma_offset_x + chroma_subblock_size_x, |
1373 | 61.4k | chroma_grain_stride, cr_col_buf, 2 >> chroma_subsamp_x, |
1374 | 61.4k | 2 >> chroma_subsamp_x, |
1375 | 61.4k | AOMMIN(chroma_subblock_size_y + (2 >> chroma_subsamp_y), |
1376 | 61.4k | (height - (y << 1)) >> chroma_subsamp_y)); |
1377 | 61.4k | } |
1378 | 81.6k | } |
1379 | 17.1k | } |
1380 | | |
1381 | 903 | dealloc_arrays(params, &pred_pos_luma, &pred_pos_chroma, &luma_grain_block, |
1382 | 903 | &cb_grain_block, &cr_grain_block, &y_line_buf, &cb_line_buf, |
1383 | 903 | &cr_line_buf, &y_col_buf, &cb_col_buf, &cr_col_buf); |
1384 | 903 | return 0; |
1385 | 903 | } |
1386 | | |
1387 | | int av1_add_film_grain(const aom_film_grain_t *params, const aom_image_t *src, |
1388 | 903 | aom_image_t *dst) { |
1389 | 903 | uint8_t *luma, *cb, *cr; |
1390 | 903 | int height, width, luma_stride, chroma_stride; |
1391 | 903 | int use_high_bit_depth = 0; |
1392 | 903 | int chroma_subsamp_x = 0; |
1393 | 903 | int chroma_subsamp_y = 0; |
1394 | 903 | int mc_identity = src->mc == AOM_CICP_MC_IDENTITY ? 1 : 0; |
1395 | | |
1396 | 903 | switch (src->fmt) { |
1397 | 198 | case AOM_IMG_FMT_I420: |
1398 | 198 | use_high_bit_depth = 0; |
1399 | 198 | chroma_subsamp_x = 1; |
1400 | 198 | chroma_subsamp_y = 1; |
1401 | 198 | break; |
1402 | 377 | case AOM_IMG_FMT_I42016: |
1403 | 377 | use_high_bit_depth = 1; |
1404 | 377 | chroma_subsamp_x = 1; |
1405 | 377 | chroma_subsamp_y = 1; |
1406 | 377 | break; |
1407 | | // case AOM_IMG_FMT_444A: |
1408 | 144 | case AOM_IMG_FMT_I444: |
1409 | 144 | use_high_bit_depth = 0; |
1410 | 144 | chroma_subsamp_x = 0; |
1411 | 144 | chroma_subsamp_y = 0; |
1412 | 144 | break; |
1413 | 130 | case AOM_IMG_FMT_I44416: |
1414 | 130 | use_high_bit_depth = 1; |
1415 | 130 | chroma_subsamp_x = 0; |
1416 | 130 | chroma_subsamp_y = 0; |
1417 | 130 | break; |
1418 | 16 | case AOM_IMG_FMT_I422: |
1419 | 16 | use_high_bit_depth = 0; |
1420 | 16 | chroma_subsamp_x = 1; |
1421 | 16 | chroma_subsamp_y = 0; |
1422 | 16 | break; |
1423 | 38 | case AOM_IMG_FMT_I42216: |
1424 | 38 | use_high_bit_depth = 1; |
1425 | 38 | chroma_subsamp_x = 1; |
1426 | 38 | chroma_subsamp_y = 0; |
1427 | 38 | break; |
1428 | 0 | default: // unknown input format |
1429 | 0 | fprintf(stderr, "Film grain error: input format is not supported!"); |
1430 | 0 | return -1; |
1431 | 903 | } |
1432 | | |
1433 | 903 | assert(params->bit_depth == src->bit_depth); |
1434 | | |
1435 | 903 | dst->fmt = src->fmt; |
1436 | 903 | dst->bit_depth = src->bit_depth; |
1437 | | |
1438 | 903 | dst->r_w = src->r_w; |
1439 | 903 | dst->r_h = src->r_h; |
1440 | 903 | dst->d_w = src->d_w; |
1441 | 903 | dst->d_h = src->d_h; |
1442 | | |
1443 | 903 | dst->cp = src->cp; |
1444 | 903 | dst->tc = src->tc; |
1445 | 903 | dst->mc = src->mc; |
1446 | | |
1447 | 903 | dst->monochrome = src->monochrome; |
1448 | 903 | dst->csp = src->csp; |
1449 | 903 | dst->range = src->range; |
1450 | | |
1451 | 903 | dst->x_chroma_shift = src->x_chroma_shift; |
1452 | 903 | dst->y_chroma_shift = src->y_chroma_shift; |
1453 | | |
1454 | 903 | dst->temporal_id = src->temporal_id; |
1455 | 903 | dst->spatial_id = src->spatial_id; |
1456 | | |
1457 | 903 | width = src->d_w % 2 ? src->d_w + 1 : src->d_w; |
1458 | 903 | height = src->d_h % 2 ? src->d_h + 1 : src->d_h; |
1459 | | |
1460 | 903 | copy_rect(src->planes[AOM_PLANE_Y], src->stride[AOM_PLANE_Y], |
1461 | 903 | dst->planes[AOM_PLANE_Y], dst->stride[AOM_PLANE_Y], src->d_w, |
1462 | 903 | src->d_h, use_high_bit_depth); |
1463 | | // Note that dst is already assumed to be aligned to even. |
1464 | 903 | extend_even(dst->planes[AOM_PLANE_Y], dst->stride[AOM_PLANE_Y], src->d_w, |
1465 | 903 | src->d_h, use_high_bit_depth); |
1466 | | |
1467 | 903 | if (!src->monochrome) { |
1468 | 557 | copy_rect(src->planes[AOM_PLANE_U], src->stride[AOM_PLANE_U], |
1469 | 557 | dst->planes[AOM_PLANE_U], dst->stride[AOM_PLANE_U], |
1470 | 557 | width >> chroma_subsamp_x, height >> chroma_subsamp_y, |
1471 | 557 | use_high_bit_depth); |
1472 | | |
1473 | 557 | copy_rect(src->planes[AOM_PLANE_V], src->stride[AOM_PLANE_V], |
1474 | 557 | dst->planes[AOM_PLANE_V], dst->stride[AOM_PLANE_V], |
1475 | 557 | width >> chroma_subsamp_x, height >> chroma_subsamp_y, |
1476 | 557 | use_high_bit_depth); |
1477 | 557 | } |
1478 | | |
1479 | 903 | luma = dst->planes[AOM_PLANE_Y]; |
1480 | 903 | cb = dst->planes[AOM_PLANE_U]; |
1481 | 903 | cr = dst->planes[AOM_PLANE_V]; |
1482 | | |
1483 | | // luma and chroma strides in samples |
1484 | 903 | luma_stride = dst->stride[AOM_PLANE_Y] >> use_high_bit_depth; |
1485 | 903 | chroma_stride = dst->stride[AOM_PLANE_U] >> use_high_bit_depth; |
1486 | | |
1487 | 903 | return add_film_grain_run(params, luma, cb, cr, height, width, luma_stride, |
1488 | 903 | chroma_stride, use_high_bit_depth, chroma_subsamp_y, |
1489 | 903 | chroma_subsamp_x, mc_identity); |
1490 | 903 | } |