1 /*
2 * Copyright (c) 2016 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10 #include <assert.h>
11 #include <stdlib.h>
12 #include "./vpx_dsp_rtcd.h"
13 #include "vpx/vpx_integer.h"
14
15 const int16_t vpx_rv[] = {
16 8, 5, 2, 2, 8, 12, 4, 9, 8, 3, 0, 3, 9, 0, 0, 0, 8, 3, 14,
17 4, 10, 1, 11, 14, 1, 14, 9, 6, 12, 11, 8, 6, 10, 0, 0, 8, 9, 0,
18 3, 14, 8, 11, 13, 4, 2, 9, 0, 3, 9, 6, 1, 2, 3, 14, 13, 1, 8,
19 2, 9, 7, 3, 3, 1, 13, 13, 6, 6, 5, 2, 7, 11, 9, 11, 8, 7, 3,
20 2, 0, 13, 13, 14, 4, 12, 5, 12, 10, 8, 10, 13, 10, 4, 14, 4, 10, 0,
21 8, 11, 1, 13, 7, 7, 14, 6, 14, 13, 2, 13, 5, 4, 4, 0, 10, 0, 5,
22 13, 2, 12, 7, 11, 13, 8, 0, 4, 10, 7, 2, 7, 2, 2, 5, 3, 4, 7,
23 3, 3, 14, 14, 5, 9, 13, 3, 14, 3, 6, 3, 0, 11, 8, 13, 1, 13, 1,
24 12, 0, 10, 9, 7, 6, 2, 8, 5, 2, 13, 7, 1, 13, 14, 7, 6, 7, 9,
25 6, 10, 11, 7, 8, 7, 5, 14, 8, 4, 4, 0, 8, 7, 10, 0, 8, 14, 11,
26 3, 12, 5, 7, 14, 3, 14, 5, 2, 6, 11, 12, 12, 8, 0, 11, 13, 1, 2,
27 0, 5, 10, 14, 7, 8, 0, 4, 11, 0, 8, 0, 3, 10, 5, 8, 0, 11, 6,
28 7, 8, 10, 7, 13, 9, 2, 5, 1, 5, 10, 2, 4, 3, 5, 6, 10, 8, 9,
29 4, 11, 14, 0, 10, 0, 5, 13, 2, 12, 7, 11, 13, 8, 0, 4, 10, 7, 2,
30 7, 2, 2, 5, 3, 4, 7, 3, 3, 14, 14, 5, 9, 13, 3, 14, 3, 6, 3,
31 0, 11, 8, 13, 1, 13, 1, 12, 0, 10, 9, 7, 6, 2, 8, 5, 2, 13, 7,
32 1, 13, 14, 7, 6, 7, 9, 6, 10, 11, 7, 8, 7, 5, 14, 8, 4, 4, 0,
33 8, 7, 10, 0, 8, 14, 11, 3, 12, 5, 7, 14, 3, 14, 5, 2, 6, 11, 12,
34 12, 8, 0, 11, 13, 1, 2, 0, 5, 10, 14, 7, 8, 0, 4, 11, 0, 8, 0,
35 3, 10, 5, 8, 0, 11, 6, 7, 8, 10, 7, 13, 9, 2, 5, 1, 5, 10, 2,
36 4, 3, 5, 6, 10, 8, 9, 4, 11, 14, 3, 8, 3, 7, 8, 5, 11, 4, 12,
37 3, 11, 9, 14, 8, 14, 13, 4, 3, 1, 2, 14, 6, 5, 4, 4, 11, 4, 6,
38 2, 1, 5, 8, 8, 12, 13, 5, 14, 10, 12, 13, 0, 9, 5, 5, 11, 10, 13,
39 9, 10, 13,
40 };
41
vpx_post_proc_down_and_across_mb_row_c(unsigned char * src,unsigned char * dst,int src_pitch,int dst_pitch,int cols,unsigned char * flimits,int size)42 void vpx_post_proc_down_and_across_mb_row_c(unsigned char *src,
43 unsigned char *dst, int src_pitch,
44 int dst_pitch, int cols,
45 unsigned char *flimits, int size) {
46 unsigned char *p_src, *p_dst;
47 int row;
48 int col;
49 unsigned char v;
50 unsigned char d[4];
51
52 assert(size >= 8);
53 assert(cols >= 8);
54
55 for (row = 0; row < size; row++) {
56 /* post_proc_down for one row */
57 p_src = src;
58 p_dst = dst;
59
60 for (col = 0; col < cols; col++) {
61 unsigned char p_above2 = p_src[col - 2 * src_pitch];
62 unsigned char p_above1 = p_src[col - src_pitch];
63 unsigned char p_below1 = p_src[col + src_pitch];
64 unsigned char p_below2 = p_src[col + 2 * src_pitch];
65
66 v = p_src[col];
67
68 if ((abs(v - p_above2) < flimits[col]) &&
69 (abs(v - p_above1) < flimits[col]) &&
70 (abs(v - p_below1) < flimits[col]) &&
71 (abs(v - p_below2) < flimits[col])) {
72 unsigned char k1, k2, k3;
73 k1 = (p_above2 + p_above1 + 1) >> 1;
74 k2 = (p_below2 + p_below1 + 1) >> 1;
75 k3 = (k1 + k2 + 1) >> 1;
76 v = (k3 + v + 1) >> 1;
77 }
78
79 p_dst[col] = v;
80 }
81
82 /* now post_proc_across */
83 p_src = dst;
84 p_dst = dst;
85
86 p_src[-2] = p_src[-1] = p_src[0];
87 p_src[cols] = p_src[cols + 1] = p_src[cols - 1];
88
89 for (col = 0; col < cols; col++) {
90 v = p_src[col];
91
92 if ((abs(v - p_src[col - 2]) < flimits[col]) &&
93 (abs(v - p_src[col - 1]) < flimits[col]) &&
94 (abs(v - p_src[col + 1]) < flimits[col]) &&
95 (abs(v - p_src[col + 2]) < flimits[col])) {
96 unsigned char k1, k2, k3;
97 k1 = (p_src[col - 2] + p_src[col - 1] + 1) >> 1;
98 k2 = (p_src[col + 2] + p_src[col + 1] + 1) >> 1;
99 k3 = (k1 + k2 + 1) >> 1;
100 v = (k3 + v + 1) >> 1;
101 }
102
103 d[col & 3] = v;
104
105 if (col >= 2) p_dst[col - 2] = d[(col - 2) & 3];
106 }
107
108 /* handle the last two pixels */
109 p_dst[col - 2] = d[(col - 2) & 3];
110 p_dst[col - 1] = d[(col - 1) & 3];
111
112 /* next row */
113 src += src_pitch;
114 dst += dst_pitch;
115 }
116 }
117
vpx_mbpost_proc_across_ip_c(unsigned char * src,int pitch,int rows,int cols,int flimit)118 void vpx_mbpost_proc_across_ip_c(unsigned char *src, int pitch, int rows,
119 int cols, int flimit) {
120 int r, c, i;
121
122 unsigned char *s = src;
123 unsigned char d[16];
124
125 for (r = 0; r < rows; r++) {
126 int sumsq = 16;
127 int sum = 0;
128
129 for (i = -8; i < 0; i++) s[i] = s[0];
130
131 /* 17 avoids valgrind warning - we buffer values in c in d
132 * and only write them when we've read 8 ahead...
133 */
134 for (i = 0; i < 17; i++) s[i + cols] = s[cols - 1];
135
136 for (i = -8; i <= 6; i++) {
137 sumsq += s[i] * s[i];
138 sum += s[i];
139 d[i + 8] = 0;
140 }
141
142 for (c = 0; c < cols + 8; c++) {
143 int x = s[c + 7] - s[c - 8];
144 int y = s[c + 7] + s[c - 8];
145
146 sum += x;
147 sumsq += x * y;
148
149 d[c & 15] = s[c];
150
151 if (sumsq * 15 - sum * sum < flimit) {
152 d[c & 15] = (8 + sum + s[c]) >> 4;
153 }
154
155 s[c - 8] = d[(c - 8) & 15];
156 }
157
158 s += pitch;
159 }
160 }
161
vpx_mbpost_proc_down_c(unsigned char * dst,int pitch,int rows,int cols,int flimit)162 void vpx_mbpost_proc_down_c(unsigned char *dst, int pitch, int rows, int cols,
163 int flimit) {
164 int r, c, i;
165
166 for (c = 0; c < cols; c++) {
167 unsigned char *s = &dst[c];
168 int sumsq = 0;
169 int sum = 0;
170 unsigned char d[16];
171
172 for (i = -8; i < 0; i++) s[i * pitch] = s[0];
173
174 /* 17 avoids valgrind warning - we buffer values in c in d
175 * and only write them when we've read 8 ahead...
176 */
177 for (i = 0; i < 17; i++) s[(i + rows) * pitch] = s[(rows - 1) * pitch];
178
179 for (i = -8; i <= 6; i++) {
180 sumsq += s[i * pitch] * s[i * pitch];
181 sum += s[i * pitch];
182 }
183
184 for (r = 0; r < rows + 8; r++) {
185 sumsq += s[7 * pitch] * s[7 * pitch] - s[-8 * pitch] * s[-8 * pitch];
186 sum += s[7 * pitch] - s[-8 * pitch];
187 d[r & 15] = s[0];
188
189 if (sumsq * 15 - sum * sum < flimit) {
190 d[r & 15] = (vpx_rv[(r & 127) + (c & 7)] + sum + s[0]) >> 4;
191 }
192 if (r >= 8) s[-8 * pitch] = d[(r - 8) & 15];
193 s += pitch;
194 }
195 }
196 }
197