1 // Copyright 2010 Google Inc. All Rights Reserved.
2 //
3 // Use of this source code is governed by a BSD-style license
4 // that can be found in the COPYING file in the root of the source
5 // tree. An additional intellectual property rights grant can be found
6 // in the file PATENTS. All contributing project authors may
7 // be found in the AUTHORS file in the root of the source tree.
8 // -----------------------------------------------------------------------------
9 //
10 // YUV->RGB conversion functions
11 //
12 // Author: Skal (pascal.massimino@gmail.com)
13
14 #include "./yuv.h"
15
16 #if defined(WEBP_YUV_USE_TABLE)
17
18 static int done = 0;
19
clip(int v,int max_value)20 static WEBP_INLINE uint8_t clip(int v, int max_value) {
21 return v < 0 ? 0 : v > max_value ? max_value : v;
22 }
23
24 int16_t VP8kVToR[256], VP8kUToB[256];
25 int32_t VP8kVToG[256], VP8kUToG[256];
26 uint8_t VP8kClip[YUV_RANGE_MAX - YUV_RANGE_MIN];
27 uint8_t VP8kClip4Bits[YUV_RANGE_MAX - YUV_RANGE_MIN];
28
VP8YUVInit(void)29 WEBP_TSAN_IGNORE_FUNCTION void VP8YUVInit(void) {
30 int i;
31 if (done) {
32 return;
33 }
34 #ifndef USE_YUVj
35 for (i = 0; i < 256; ++i) {
36 VP8kVToR[i] = (89858 * (i - 128) + YUV_HALF) >> YUV_FIX;
37 VP8kUToG[i] = -22014 * (i - 128) + YUV_HALF;
38 VP8kVToG[i] = -45773 * (i - 128);
39 VP8kUToB[i] = (113618 * (i - 128) + YUV_HALF) >> YUV_FIX;
40 }
41 for (i = YUV_RANGE_MIN; i < YUV_RANGE_MAX; ++i) {
42 const int k = ((i - 16) * 76283 + YUV_HALF) >> YUV_FIX;
43 VP8kClip[i - YUV_RANGE_MIN] = clip(k, 255);
44 VP8kClip4Bits[i - YUV_RANGE_MIN] = clip((k + 8) >> 4, 15);
45 }
46 #else
47 for (i = 0; i < 256; ++i) {
48 VP8kVToR[i] = (91881 * (i - 128) + YUV_HALF) >> YUV_FIX;
49 VP8kUToG[i] = -22554 * (i - 128) + YUV_HALF;
50 VP8kVToG[i] = -46802 * (i - 128);
51 VP8kUToB[i] = (116130 * (i - 128) + YUV_HALF) >> YUV_FIX;
52 }
53 for (i = YUV_RANGE_MIN; i < YUV_RANGE_MAX; ++i) {
54 const int k = i;
55 VP8kClip[i - YUV_RANGE_MIN] = clip(k, 255);
56 VP8kClip4Bits[i - YUV_RANGE_MIN] = clip((k + 8) >> 4, 15);
57 }
58 #endif
59
60 done = 1;
61 }
62
63 #else
64
VP8YUVInit(void)65 WEBP_TSAN_IGNORE_FUNCTION void VP8YUVInit(void) {}
66
67 #endif // WEBP_YUV_USE_TABLE
68
69 //-----------------------------------------------------------------------------
70 // Plain-C version
71
72 #define ROW_FUNC(FUNC_NAME, FUNC, XSTEP) \
73 static void FUNC_NAME(const uint8_t* y, \
74 const uint8_t* u, const uint8_t* v, \
75 uint8_t* dst, int len) { \
76 const uint8_t* const end = dst + (len & ~1) * XSTEP; \
77 while (dst != end) { \
78 FUNC(y[0], u[0], v[0], dst); \
79 FUNC(y[1], u[0], v[0], dst + XSTEP); \
80 y += 2; \
81 ++u; \
82 ++v; \
83 dst += 2 * XSTEP; \
84 } \
85 if (len & 1) { \
86 FUNC(y[0], u[0], v[0], dst); \
87 } \
88 } \
89
90 // All variants implemented.
91 ROW_FUNC(YuvToRgbRow, VP8YuvToRgb, 3)
92 ROW_FUNC(YuvToBgrRow, VP8YuvToBgr, 3)
93 ROW_FUNC(YuvToRgbaRow, VP8YuvToRgba, 4)
94 ROW_FUNC(YuvToBgraRow, VP8YuvToBgra, 4)
95 ROW_FUNC(YuvToArgbRow, VP8YuvToArgb, 4)
96 ROW_FUNC(YuvToRgba4444Row, VP8YuvToRgba4444, 2)
97 ROW_FUNC(YuvToRgb565Row, VP8YuvToRgb565, 2)
98
99 #undef ROW_FUNC
100
101 // Main call for processing a plane with a WebPSamplerRowFunc function:
WebPSamplerProcessPlane(const uint8_t * y,int y_stride,const uint8_t * u,const uint8_t * v,int uv_stride,uint8_t * dst,int dst_stride,int width,int height,WebPSamplerRowFunc func)102 void WebPSamplerProcessPlane(const uint8_t* y, int y_stride,
103 const uint8_t* u, const uint8_t* v, int uv_stride,
104 uint8_t* dst, int dst_stride,
105 int width, int height, WebPSamplerRowFunc func) {
106 int j;
107 for (j = 0; j < height; ++j) {
108 func(y, u, v, dst, width);
109 y += y_stride;
110 if (j & 1) {
111 u += uv_stride;
112 v += uv_stride;
113 }
114 dst += dst_stride;
115 }
116 }
117
118 //-----------------------------------------------------------------------------
119 // Main call
120
121 WebPSamplerRowFunc WebPSamplers[MODE_LAST];
122
123 extern void WebPInitSamplersSSE2(void);
124 extern void WebPInitSamplersMIPS32(void);
125 extern void WebPInitSamplersMIPSdspR2(void);
126
127 static volatile VP8CPUInfo yuv_last_cpuinfo_used =
128 (VP8CPUInfo)&yuv_last_cpuinfo_used;
129
WebPInitSamplers(void)130 WEBP_TSAN_IGNORE_FUNCTION void WebPInitSamplers(void) {
131 if (yuv_last_cpuinfo_used == VP8GetCPUInfo) return;
132
133 WebPSamplers[MODE_RGB] = YuvToRgbRow;
134 WebPSamplers[MODE_RGBA] = YuvToRgbaRow;
135 WebPSamplers[MODE_BGR] = YuvToBgrRow;
136 WebPSamplers[MODE_BGRA] = YuvToBgraRow;
137 WebPSamplers[MODE_ARGB] = YuvToArgbRow;
138 WebPSamplers[MODE_RGBA_4444] = YuvToRgba4444Row;
139 WebPSamplers[MODE_RGB_565] = YuvToRgb565Row;
140 WebPSamplers[MODE_rgbA] = YuvToRgbaRow;
141 WebPSamplers[MODE_bgrA] = YuvToBgraRow;
142 WebPSamplers[MODE_Argb] = YuvToArgbRow;
143 WebPSamplers[MODE_rgbA_4444] = YuvToRgba4444Row;
144
145 // If defined, use CPUInfo() to overwrite some pointers with faster versions.
146 if (VP8GetCPUInfo != NULL) {
147 #if defined(WEBP_USE_SSE2)
148 if (VP8GetCPUInfo(kSSE2)) {
149 WebPInitSamplersSSE2();
150 }
151 #endif // WEBP_USE_SSE2
152 #if defined(WEBP_USE_MIPS32)
153 if (VP8GetCPUInfo(kMIPS32)) {
154 WebPInitSamplersMIPS32();
155 }
156 #endif // WEBP_USE_MIPS32
157 #if defined(WEBP_USE_MIPS_DSP_R2)
158 if (VP8GetCPUInfo(kMIPSdspR2)) {
159 WebPInitSamplersMIPSdspR2();
160 }
161 #endif // WEBP_USE_MIPS_DSP_R2
162 }
163 yuv_last_cpuinfo_used = VP8GetCPUInfo;
164 }
165
166 //-----------------------------------------------------------------------------
167 // ARGB -> YUV converters
168
ConvertARGBToY(const uint32_t * argb,uint8_t * y,int width)169 static void ConvertARGBToY(const uint32_t* argb, uint8_t* y, int width) {
170 int i;
171 for (i = 0; i < width; ++i) {
172 const uint32_t p = argb[i];
173 y[i] = VP8RGBToY((p >> 16) & 0xff, (p >> 8) & 0xff, (p >> 0) & 0xff,
174 YUV_HALF);
175 }
176 }
177
WebPConvertARGBToUV_C(const uint32_t * argb,uint8_t * u,uint8_t * v,int src_width,int do_store)178 void WebPConvertARGBToUV_C(const uint32_t* argb, uint8_t* u, uint8_t* v,
179 int src_width, int do_store) {
180 // No rounding. Last pixel is dealt with separately.
181 const int uv_width = src_width >> 1;
182 int i;
183 for (i = 0; i < uv_width; ++i) {
184 const uint32_t v0 = argb[2 * i + 0];
185 const uint32_t v1 = argb[2 * i + 1];
186 // VP8RGBToU/V expects four accumulated pixels. Hence we need to
187 // scale r/g/b value by a factor 2. We just shift v0/v1 one bit less.
188 const int r = ((v0 >> 15) & 0x1fe) + ((v1 >> 15) & 0x1fe);
189 const int g = ((v0 >> 7) & 0x1fe) + ((v1 >> 7) & 0x1fe);
190 const int b = ((v0 << 1) & 0x1fe) + ((v1 << 1) & 0x1fe);
191 const int tmp_u = VP8RGBToU(r, g, b, YUV_HALF << 2);
192 const int tmp_v = VP8RGBToV(r, g, b, YUV_HALF << 2);
193 if (do_store) {
194 u[i] = tmp_u;
195 v[i] = tmp_v;
196 } else {
197 // Approximated average-of-four. But it's an acceptable diff.
198 u[i] = (u[i] + tmp_u + 1) >> 1;
199 v[i] = (v[i] + tmp_v + 1) >> 1;
200 }
201 }
202 if (src_width & 1) { // last pixel
203 const uint32_t v0 = argb[2 * i + 0];
204 const int r = (v0 >> 14) & 0x3fc;
205 const int g = (v0 >> 6) & 0x3fc;
206 const int b = (v0 << 2) & 0x3fc;
207 const int tmp_u = VP8RGBToU(r, g, b, YUV_HALF << 2);
208 const int tmp_v = VP8RGBToV(r, g, b, YUV_HALF << 2);
209 if (do_store) {
210 u[i] = tmp_u;
211 v[i] = tmp_v;
212 } else {
213 u[i] = (u[i] + tmp_u + 1) >> 1;
214 v[i] = (v[i] + tmp_v + 1) >> 1;
215 }
216 }
217 }
218
219 //-----------------------------------------------------------------------------
220
ConvertRGB24ToY(const uint8_t * rgb,uint8_t * y,int width)221 static void ConvertRGB24ToY(const uint8_t* rgb, uint8_t* y, int width) {
222 int i;
223 for (i = 0; i < width; ++i, rgb += 3) {
224 y[i] = VP8RGBToY(rgb[0], rgb[1], rgb[2], YUV_HALF);
225 }
226 }
227
ConvertBGR24ToY(const uint8_t * bgr,uint8_t * y,int width)228 static void ConvertBGR24ToY(const uint8_t* bgr, uint8_t* y, int width) {
229 int i;
230 for (i = 0; i < width; ++i, bgr += 3) {
231 y[i] = VP8RGBToY(bgr[2], bgr[1], bgr[0], YUV_HALF);
232 }
233 }
234
WebPConvertRGBA32ToUV_C(const uint16_t * rgb,uint8_t * u,uint8_t * v,int width)235 void WebPConvertRGBA32ToUV_C(const uint16_t* rgb,
236 uint8_t* u, uint8_t* v, int width) {
237 int i;
238 for (i = 0; i < width; i += 1, rgb += 4) {
239 const int r = rgb[0], g = rgb[1], b = rgb[2];
240 u[i] = VP8RGBToU(r, g, b, YUV_HALF << 2);
241 v[i] = VP8RGBToV(r, g, b, YUV_HALF << 2);
242 }
243 }
244
245 //-----------------------------------------------------------------------------
246
247 void (*WebPConvertRGB24ToY)(const uint8_t* rgb, uint8_t* y, int width);
248 void (*WebPConvertBGR24ToY)(const uint8_t* bgr, uint8_t* y, int width);
249 void (*WebPConvertRGBA32ToUV)(const uint16_t* rgb,
250 uint8_t* u, uint8_t* v, int width);
251
252 void (*WebPConvertARGBToY)(const uint32_t* argb, uint8_t* y, int width);
253 void (*WebPConvertARGBToUV)(const uint32_t* argb, uint8_t* u, uint8_t* v,
254 int src_width, int do_store);
255
256 static volatile VP8CPUInfo rgba_to_yuv_last_cpuinfo_used =
257 (VP8CPUInfo)&rgba_to_yuv_last_cpuinfo_used;
258
259 extern void WebPInitConvertARGBToYUVSSE2(void);
260
WebPInitConvertARGBToYUV(void)261 WEBP_TSAN_IGNORE_FUNCTION void WebPInitConvertARGBToYUV(void) {
262 if (rgba_to_yuv_last_cpuinfo_used == VP8GetCPUInfo) return;
263
264 WebPConvertARGBToY = ConvertARGBToY;
265 WebPConvertARGBToUV = WebPConvertARGBToUV_C;
266
267 WebPConvertRGB24ToY = ConvertRGB24ToY;
268 WebPConvertBGR24ToY = ConvertBGR24ToY;
269
270 WebPConvertRGBA32ToUV = WebPConvertRGBA32ToUV_C;
271
272 if (VP8GetCPUInfo != NULL) {
273 #if defined(WEBP_USE_SSE2)
274 if (VP8GetCPUInfo(kSSE2)) {
275 WebPInitConvertARGBToYUVSSE2();
276 }
277 #endif // WEBP_USE_SSE2
278 }
279 rgba_to_yuv_last_cpuinfo_used = VP8GetCPUInfo;
280 }
281