/src/vvenc/source/Lib/CommonLib/SampleAdaptiveOffset.cpp
Line | Count | Source |
1 | | /* ----------------------------------------------------------------------------- |
2 | | The copyright in this software is being made available under the Clear BSD |
3 | | License, included below. No patent rights, trademark rights and/or |
4 | | other Intellectual Property Rights other than the copyrights concerning |
5 | | the Software are granted under this license. |
6 | | |
7 | | The Clear BSD License |
8 | | |
9 | | Copyright (c) 2019-2026, Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. & The VVenC Authors. |
10 | | All rights reserved. |
11 | | |
12 | | Redistribution and use in source and binary forms, with or without modification, |
13 | | are permitted (subject to the limitations in the disclaimer below) provided that |
14 | | the following conditions are met: |
15 | | |
16 | | * Redistributions of source code must retain the above copyright notice, |
17 | | this list of conditions and the following disclaimer. |
18 | | |
19 | | * Redistributions in binary form must reproduce the above copyright |
20 | | notice, this list of conditions and the following disclaimer in the |
21 | | documentation and/or other materials provided with the distribution. |
22 | | |
23 | | * Neither the name of the copyright holder nor the names of its |
24 | | contributors may be used to endorse or promote products derived from this |
25 | | software without specific prior written permission. |
26 | | |
27 | | NO EXPRESS OR IMPLIED LICENSES TO ANY PARTY'S PATENT RIGHTS ARE GRANTED BY |
28 | | THIS LICENSE. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND |
29 | | CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
30 | | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A |
31 | | PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR |
32 | | CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
33 | | EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
34 | | PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR |
35 | | BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER |
36 | | IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
37 | | ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
38 | | POSSIBILITY OF SUCH DAMAGE. |
39 | | |
40 | | |
41 | | ------------------------------------------------------------------------------------------- */ |
42 | | |
43 | | |
44 | | /** \file SampleAdaptiveOffset.cpp |
45 | | \brief sample adaptive offset class |
46 | | */ |
47 | | |
48 | | #include "SampleAdaptiveOffset.h" |
49 | | #include "UnitTools.h" |
50 | | #include "UnitPartitioner.h" |
51 | | #include "CodingStructure.h" |
52 | | #include "dtrace_codingstruct.h" |
53 | | #include "dtrace_buffer.h" |
54 | | |
55 | | #include <string.h> |
56 | | #include <stdlib.h> |
57 | | #include <math.h> |
58 | | |
59 | | //! \ingroup CommonLib |
60 | | //! \{ |
61 | | |
62 | | namespace vvenc { |
63 | | |
64 | | void offsetBlock_core(const int channelBitDepth, const ClpRng& clpRng, int typeIdx, int* offset, int startIdx, |
65 | | const Pel* srcBlk, Pel* resBlk, ptrdiff_t srcStride, ptrdiff_t resStride, int width, int height, |
66 | | uint8_t availMask, std::vector<int8_t> &signLineBuf1, std::vector<int8_t> &signLineBuf2) |
67 | 0 | { |
68 | 0 | int x, y, startX, startY, endX, endY, edgeType; |
69 | 0 | int firstLineStartX, firstLineEndX, lastLineStartX, lastLineEndX; |
70 | 0 | int8_t signLeft, signRight, signDown; |
71 | |
|
72 | 0 | const Pel* srcLine = srcBlk; |
73 | 0 | Pel* resLine = resBlk; |
74 | |
|
75 | 0 | switch (typeIdx) |
76 | 0 | { |
77 | 0 | case SAO_TYPE_EO_0: |
78 | 0 | { |
79 | 0 | offset += 2; |
80 | 0 | startX = availMask&LeftAvail ? 0 : 1; |
81 | 0 | endX = availMask&RightAvail ? width : (width - 1); |
82 | 0 | for (y = 0; y < height; y++) |
83 | 0 | { |
84 | 0 | signLeft = (int8_t)sgn(srcLine[startX] - srcLine[startX - 1]); |
85 | 0 | for (x = startX; x < endX; x++) |
86 | 0 | { |
87 | 0 | signRight = (int8_t)sgn(srcLine[x] - srcLine[x + 1]); |
88 | 0 | edgeType = signRight + signLeft; |
89 | 0 | signLeft = -signRight; |
90 | |
|
91 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
92 | 0 | } |
93 | 0 | srcLine += srcStride; |
94 | 0 | resLine += resStride; |
95 | 0 | } |
96 | |
|
97 | 0 | } |
98 | 0 | break; |
99 | 0 | case SAO_TYPE_EO_90: |
100 | 0 | { |
101 | 0 | offset += 2; |
102 | 0 | int8_t *signUpLine = &signLineBuf1[0]; |
103 | |
|
104 | 0 | startY = availMask&AboveAvail ? 0 : 1; |
105 | 0 | endY = availMask&BelowAvail ? height : height - 1; |
106 | 0 | if (!(availMask&AboveAvail)) |
107 | 0 | { |
108 | 0 | srcLine += srcStride; |
109 | 0 | resLine += resStride; |
110 | 0 | } |
111 | |
|
112 | 0 | const Pel* srcLineAbove = srcLine - srcStride; |
113 | 0 | for (x = 0; x < width; x++) |
114 | 0 | { |
115 | 0 | signUpLine[x] = (int8_t)sgn(srcLine[x] - srcLineAbove[x]); |
116 | 0 | } |
117 | |
|
118 | 0 | const Pel* srcLineBelow; |
119 | 0 | for (y = startY; y < endY; y++) |
120 | 0 | { |
121 | 0 | srcLineBelow = srcLine + srcStride; |
122 | |
|
123 | 0 | for (x = 0; x < width; x++) |
124 | 0 | { |
125 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x]); |
126 | 0 | edgeType = signDown + signUpLine[x]; |
127 | 0 | signUpLine[x] = -signDown; |
128 | |
|
129 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
130 | 0 | } |
131 | 0 | srcLine += srcStride; |
132 | 0 | resLine += resStride; |
133 | 0 | } |
134 | |
|
135 | 0 | } |
136 | 0 | break; |
137 | 0 | case SAO_TYPE_EO_135: |
138 | 0 | { |
139 | 0 | offset += 2; |
140 | 0 | int8_t *signUpLine, *signDownLine, *signTmpLine; |
141 | |
|
142 | 0 | signUpLine = &signLineBuf1[0]; |
143 | 0 | signDownLine = &signLineBuf2[0]; |
144 | |
|
145 | 0 | startX = availMask&LeftAvail ? 0 : 1; |
146 | 0 | endX = availMask&RightAvail ? width : (width - 1); |
147 | | |
148 | | //prepare 2nd line's upper sign |
149 | 0 | const Pel* srcLineBelow = srcLine + srcStride; |
150 | 0 | for (x = startX; x < endX + 1; x++) |
151 | 0 | { |
152 | 0 | signUpLine[x] = (int8_t)sgn(srcLineBelow[x] - srcLine[x - 1]); |
153 | 0 | } |
154 | | |
155 | | //1st line |
156 | 0 | const Pel* srcLineAbove = srcLine - srcStride; |
157 | 0 | firstLineStartX = availMask&AboveLeftAvail ? 0 : 1; |
158 | 0 | firstLineEndX = availMask&AboveAvail ? endX : 1; |
159 | 0 | for (x = firstLineStartX; x < firstLineEndX; x++) |
160 | 0 | { |
161 | 0 | edgeType = sgn(srcLine[x] - srcLineAbove[x - 1]) - signUpLine[x + 1]; |
162 | |
|
163 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
164 | 0 | } |
165 | 0 | srcLine += srcStride; |
166 | 0 | resLine += resStride; |
167 | | |
168 | | |
169 | | //middle lines |
170 | 0 | for (y = 1; y < height - 1; y++) |
171 | 0 | { |
172 | 0 | srcLineBelow = srcLine + srcStride; |
173 | |
|
174 | 0 | for (x = startX; x < endX; x++) |
175 | 0 | { |
176 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x + 1]); |
177 | 0 | edgeType = signDown + signUpLine[x]; |
178 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
179 | |
|
180 | 0 | signDownLine[x + 1] = -signDown; |
181 | 0 | } |
182 | 0 | signDownLine[startX] = (int8_t)sgn(srcLineBelow[startX] - srcLine[startX - 1]); |
183 | |
|
184 | 0 | signTmpLine = signUpLine; |
185 | 0 | signUpLine = signDownLine; |
186 | 0 | signDownLine = signTmpLine; |
187 | |
|
188 | 0 | srcLine += srcStride; |
189 | 0 | resLine += resStride; |
190 | 0 | } |
191 | | |
192 | | //last line |
193 | 0 | srcLineBelow = srcLine + srcStride; |
194 | 0 | lastLineStartX = availMask&BelowAvail ? startX : (width - 1); |
195 | 0 | lastLineEndX = availMask&BelowRightAvail ? width : (width - 1); |
196 | 0 | for (x = lastLineStartX; x < lastLineEndX; x++) |
197 | 0 | { |
198 | 0 | edgeType = sgn(srcLine[x] - srcLineBelow[x + 1]) + signUpLine[x]; |
199 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
200 | |
|
201 | 0 | } |
202 | 0 | } |
203 | 0 | break; |
204 | 0 | case SAO_TYPE_EO_45: |
205 | 0 | { |
206 | 0 | offset += 2; |
207 | 0 | int8_t *signUpLine = &signLineBuf1[1]; |
208 | |
|
209 | 0 | startX = availMask&LeftAvail ? 0 : 1; |
210 | 0 | endX = availMask&RightAvail ? width : (width - 1); |
211 | | |
212 | | //prepare 2nd line upper sign |
213 | 0 | const Pel* srcLineBelow = srcLine + srcStride; |
214 | 0 | for (x = startX - 1; x < endX; x++) |
215 | 0 | { |
216 | 0 | signUpLine[x] = (int8_t)sgn(srcLineBelow[x] - srcLine[x + 1]); |
217 | 0 | } |
218 | | |
219 | | |
220 | | //first line |
221 | 0 | const Pel* srcLineAbove = srcLine - srcStride; |
222 | 0 | firstLineStartX = availMask&AboveAvail ? startX : (width - 1); |
223 | 0 | firstLineEndX = availMask&AboveRightAvail ? width : (width - 1); |
224 | 0 | for (x = firstLineStartX; x < firstLineEndX; x++) |
225 | 0 | { |
226 | 0 | edgeType = sgn(srcLine[x] - srcLineAbove[x + 1]) - signUpLine[x - 1]; |
227 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
228 | 0 | } |
229 | 0 | srcLine += srcStride; |
230 | 0 | resLine += resStride; |
231 | | |
232 | | //middle lines |
233 | 0 | for (y = 1; y < height - 1; y++) |
234 | 0 | { |
235 | 0 | srcLineBelow = srcLine + srcStride; |
236 | |
|
237 | 0 | for (x = startX; x < endX; x++) |
238 | 0 | { |
239 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x - 1]); |
240 | 0 | edgeType = signDown + signUpLine[x]; |
241 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
242 | 0 | signUpLine[x - 1] = -signDown; |
243 | 0 | } |
244 | 0 | signUpLine[endX - 1] = (int8_t)sgn(srcLineBelow[endX - 1] - srcLine[endX]); |
245 | 0 | srcLine += srcStride; |
246 | 0 | resLine += resStride; |
247 | 0 | } |
248 | | |
249 | | //last line |
250 | 0 | srcLineBelow = srcLine + srcStride; |
251 | 0 | lastLineStartX = availMask&BelowLeftAvail ? 0 : 1; |
252 | 0 | lastLineEndX = availMask&BelowAvail ? endX : 1; |
253 | 0 | for (x = lastLineStartX; x < lastLineEndX; x++) |
254 | 0 | { |
255 | 0 | edgeType = sgn(srcLine[x] - srcLineBelow[x - 1]) + signUpLine[x]; |
256 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[edgeType], clpRng); |
257 | |
|
258 | 0 | } |
259 | 0 | } |
260 | 0 | break; |
261 | 0 | case SAO_TYPE_BO: |
262 | 0 | { |
263 | 0 | const int shiftBits = channelBitDepth - NUM_SAO_BO_CLASSES_LOG2; |
264 | 0 | for (y = 0; y < height; y++) |
265 | 0 | { |
266 | 0 | for (x = 0; x < width; x++) |
267 | 0 | { |
268 | 0 | resLine[x] = ClipPel<int>(srcLine[x] + offset[srcLine[x] >> shiftBits], clpRng); |
269 | 0 | } |
270 | 0 | srcLine += srcStride; |
271 | 0 | resLine += resStride; |
272 | 0 | } |
273 | 0 | } |
274 | 0 | break; |
275 | 0 | default: |
276 | 0 | { |
277 | 0 | THROW("Not a supported SAO types\n"); |
278 | 0 | } |
279 | 0 | } |
280 | 0 | } |
281 | | void calcSaoStatisticsEo0_Core(int width,int startX,int endX,int endY,Pel* srcLine,Pel* orgLine,int srcStride,int orgStride,int64_t *count, int64_t *diff) |
282 | 0 | { |
283 | 0 | int iNaRight=width-endX; |
284 | 0 | srcLine = srcLine + startX; |
285 | 0 | orgLine =orgLine + startX; |
286 | 0 | int iNaWidth = startX + iNaRight; |
287 | 0 | int i,j; |
288 | 0 | diff +=2; |
289 | 0 | count+=2; |
290 | 0 | for ( i = 0; i < endY; i++ ) |
291 | 0 | { |
292 | 0 | int iSignLeft = sgn( *srcLine - *(srcLine - 1) ); |
293 | 0 | for ( j = 0; j < width - iNaWidth; j++, srcLine++, orgLine++ ) |
294 | 0 | { |
295 | 0 | int iSignRight = sgn( *srcLine - *(srcLine + 1) ); |
296 | 0 | int iType = iSignLeft + iSignRight; |
297 | 0 | iSignLeft = -1 * iSignRight; |
298 | 0 | diff[iType] += (*orgLine - *srcLine); |
299 | 0 | count[iType] += 1; |
300 | 0 | } |
301 | 0 | srcLine += srcStride - ( width - iNaWidth ); |
302 | 0 | orgLine += orgStride - ( width - iNaWidth ); |
303 | 0 | } |
304 | 0 | } |
305 | | void calcSaoStatisticsEo90_Core(int width,int endX,int startY,int endY,Pel* srcLine,Pel* orgLine,int srcStride,int orgStride,int64_t *count, int64_t *diff,int8_t *signUpLine) |
306 | 0 | { |
307 | 0 | diff +=2; |
308 | 0 | count+=2; |
309 | 0 | int x,y,edgeType; |
310 | 0 | Pel* srcLineAbove = srcLine - srcStride; |
311 | 0 | int8_t signDown; |
312 | 0 | for (x=0; x<endX; x++) |
313 | 0 | { |
314 | 0 | signUpLine[x] = (int8_t)sgn(srcLine[x] - srcLineAbove[x]); |
315 | 0 | } |
316 | 0 | Pel* srcLineBelow; |
317 | 0 | for (y=startY; y<endY; y++) |
318 | 0 | { |
319 | 0 | srcLineBelow = srcLine + srcStride; |
320 | 0 | for (x=0; x<endX; x++) |
321 | 0 | { |
322 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x]); |
323 | 0 | edgeType = signDown + signUpLine[x]; |
324 | 0 | signUpLine[x]= -signDown; |
325 | 0 | diff [edgeType] += (orgLine[x] - srcLine[x]); |
326 | 0 | count[edgeType] ++; |
327 | 0 | } |
328 | 0 | srcLine += srcStride; |
329 | 0 | orgLine += orgStride; |
330 | 0 | } |
331 | 0 | } |
332 | | |
333 | | |
334 | | void calcSaoStatisticsEo135_Core(int width,int startX,int endX,int endY,Pel* srcLine,Pel* orgLine,int srcStride,int orgStride,int64_t *count, int64_t *diff,int8_t *signUpLine,int8_t *signDownLine) |
335 | 0 | { |
336 | 0 | int x,y,edgeType; |
337 | 0 | int8_t signDown; |
338 | 0 | int8_t *signTmpLine; |
339 | 0 | Pel* srcLineBelow = srcLine + srcStride; |
340 | | //middle lines |
341 | 0 | for (y=1; y<endY; y++) |
342 | 0 | { |
343 | 0 | srcLineBelow = srcLine + srcStride; |
344 | 0 | for (x=startX; x<endX; x++) |
345 | 0 | { |
346 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x+1]); |
347 | 0 | edgeType = signDown + signUpLine[x]; |
348 | 0 | diff [edgeType] += (orgLine[x] - srcLine[x]); |
349 | 0 | count[edgeType] ++; |
350 | 0 | signDownLine[x+1] = -signDown; |
351 | 0 | } |
352 | 0 | signDownLine[startX] = (int8_t)sgn(srcLineBelow[startX] - srcLine[startX-1]); |
353 | 0 | signTmpLine = signUpLine; |
354 | 0 | signUpLine = signDownLine; |
355 | 0 | signDownLine = signTmpLine; |
356 | 0 | srcLine += srcStride; |
357 | 0 | orgLine += orgStride; |
358 | 0 | } |
359 | 0 | } |
360 | | void calcSaoStatisticsEo45_Core(int width,int startX,int endX,int endY,Pel* srcLine,Pel* orgLine,int srcStride,int orgStride,int64_t *count, int64_t *diff,int8_t *signUpLine) |
361 | 0 | { |
362 | 0 | int x,y,edgeType; |
363 | 0 | int8_t signDown; |
364 | 0 | Pel* srcLineBelow = srcLine + srcStride; |
365 | | //middle lines |
366 | 0 | for (y=1; y<endY; y++) |
367 | 0 | { |
368 | 0 | srcLineBelow = srcLine + srcStride; |
369 | |
|
370 | 0 | for(x=startX; x<endX; x++) |
371 | 0 | { |
372 | 0 | signDown = (int8_t)sgn(srcLine[x] - srcLineBelow[x-1]); |
373 | 0 | edgeType = signDown + signUpLine[x]; |
374 | 0 | diff [edgeType] += (orgLine[x] - srcLine[x]); |
375 | 0 | count[edgeType] ++; |
376 | 0 | signUpLine[x-1] = -signDown; |
377 | 0 | } |
378 | 0 | signUpLine[endX-1] = (int8_t)sgn(srcLineBelow[endX-1] - srcLine[endX]); |
379 | 0 | srcLine += srcStride; |
380 | 0 | orgLine += orgStride; |
381 | 0 | } |
382 | 0 | } |
383 | | void calcSaoStatisticsBo_Core(int width,int endX,int endY,Pel* srcLine,Pel* orgLine,int srcStride,int orgStride,int channelBitDepth, int64_t *count,int64_t *diff) |
384 | 0 | { |
385 | 0 | int x,y; |
386 | 0 | int startX=0; |
387 | 0 | int shiftBits = channelBitDepth - NUM_SAO_BO_CLASSES_LOG2; |
388 | 0 | for (y=0; y< endY; y++) |
389 | 0 | { |
390 | 0 | for (x=startX; x< endX; x++) |
391 | 0 | { |
392 | 0 | int bandIdx= srcLine[x] >> shiftBits; |
393 | 0 | diff [bandIdx] += (orgLine[x] - srcLine[x]); |
394 | 0 | count[bandIdx] ++; |
395 | 0 | } |
396 | 0 | srcLine += srcStride; |
397 | 0 | orgLine += orgStride; |
398 | 0 | } |
399 | 0 | } |
400 | | |
401 | | void SAOOffset::reset() |
402 | 0 | { |
403 | 0 | modeIdc = SAO_MODE_OFF; |
404 | 0 | typeIdc = -1; |
405 | 0 | typeAuxInfo = -1; |
406 | 0 | ::memset(offset, 0, sizeof(int)* MAX_NUM_SAO_CLASSES); |
407 | 0 | } |
408 | | |
409 | | void SAOBlkParam::reset() |
410 | 0 | { |
411 | 0 | for(int compIdx = 0; compIdx < MAX_NUM_COMP; compIdx++) |
412 | 0 | { |
413 | 0 | SAOOffsets[compIdx].reset(); |
414 | 0 | } |
415 | 0 | } |
416 | | |
417 | | SampleAdaptiveOffset::SampleAdaptiveOffset( bool enableOpt ) |
418 | 0 | { |
419 | 0 | offsetBlock = offsetBlock_core; |
420 | 0 | calcSaoStatisticsEo90 = calcSaoStatisticsEo90_Core; |
421 | 0 | calcSaoStatisticsEo135 = calcSaoStatisticsEo135_Core; |
422 | 0 | calcSaoStatisticsEo45 = calcSaoStatisticsEo45_Core; |
423 | 0 | calcSaoStatisticsEo0 = calcSaoStatisticsEo0_Core; |
424 | 0 | calcSaoStatisticsBo = calcSaoStatisticsBo_Core; |
425 | |
|
426 | 0 | if( enableOpt ) |
427 | 0 | { |
428 | 0 | #if ENABLE_SIMD_OPT_SAO && defined( TARGET_SIMD_X86 ) |
429 | 0 | initSampleAdaptiveOffsetX86(); |
430 | 0 | #endif |
431 | | #if ENABLE_SIMD_OPT_SAO && defined( TARGET_SIMD_ARM ) |
432 | | initSampleAdaptiveOffsetARM(); |
433 | | #endif |
434 | 0 | } |
435 | 0 | } |
436 | | |
437 | | |
438 | | SampleAdaptiveOffset::~SampleAdaptiveOffset() |
439 | 0 | { |
440 | 0 | m_signLineBuf1.clear(); |
441 | 0 | m_signLineBuf2.clear(); |
442 | 0 | } |
443 | | |
444 | | void SampleAdaptiveOffset::init( ChromaFormat format, uint32_t maxCUWidth, uint32_t maxCUHeight, uint32_t lumaBitShift, uint32_t chromaBitShift ) |
445 | 0 | { |
446 | | //bit-depth related |
447 | 0 | for(int compIdx = 0; compIdx < MAX_NUM_COMP; compIdx++) |
448 | 0 | { |
449 | 0 | m_offsetStepLog2 [compIdx] = isLuma(ComponentID(compIdx))? lumaBitShift : chromaBitShift; |
450 | 0 | } |
451 | 0 | m_numberOfComponents = getNumberValidComponents(format); |
452 | |
|
453 | 0 | size_t lineBufferSize = std::max( maxCUWidth, maxCUHeight ) + 1; |
454 | 0 | if( m_signLineBuf1.size() < lineBufferSize ) |
455 | 0 | { |
456 | 0 | m_signLineBuf1.resize( lineBufferSize ); |
457 | 0 | m_signLineBuf2.resize( lineBufferSize ); |
458 | 0 | } |
459 | 0 | } |
460 | | |
461 | | void SampleAdaptiveOffset::invertQuantOffsets(ComponentID compIdx, int typeIdc, int typeAuxInfo, int* dstOffsets, int* srcOffsets) |
462 | 0 | { |
463 | 0 | int codedOffset[MAX_NUM_SAO_CLASSES]; |
464 | |
|
465 | 0 | ::memcpy(codedOffset, srcOffsets, sizeof(int)*MAX_NUM_SAO_CLASSES); |
466 | 0 | ::memset(dstOffsets, 0, sizeof(int)*MAX_NUM_SAO_CLASSES); |
467 | |
|
468 | 0 | if(typeIdc == SAO_TYPE_START_BO) |
469 | 0 | { |
470 | 0 | for(int i=0; i< 4; i++) |
471 | 0 | { |
472 | 0 | dstOffsets[(typeAuxInfo+ i)%NUM_SAO_BO_CLASSES] = codedOffset[(typeAuxInfo+ i)%NUM_SAO_BO_CLASSES]*(1<<m_offsetStepLog2[compIdx]); |
473 | 0 | } |
474 | 0 | } |
475 | 0 | else //EO |
476 | 0 | { |
477 | 0 | for(int i=0; i< NUM_SAO_EO_CLASSES; i++) |
478 | 0 | { |
479 | 0 | dstOffsets[i] = codedOffset[i] *(1<<m_offsetStepLog2[compIdx]); |
480 | 0 | } |
481 | 0 | CHECK(dstOffsets[SAO_CLASS_EO_PLAIN] != 0, "EO offset is not '0'"); //keep EO plain offset as zero |
482 | 0 | } |
483 | |
|
484 | 0 | } |
485 | | |
486 | | int SampleAdaptiveOffset::getMergeList(CodingStructure& cs, int ctuRsAddr, SAOBlkParam* blkParams, SAOBlkParam* mergeList[NUM_SAO_MERGE_TYPES]) |
487 | 0 | { |
488 | 0 | const PreCalcValues& pcv = *cs.pcv; |
489 | |
|
490 | 0 | int ctuX = ctuRsAddr % pcv.widthInCtus; |
491 | 0 | int ctuY = ctuRsAddr / pcv.widthInCtus; |
492 | 0 | const CodingUnit& cu = *cs.getCU(Position(ctuX*pcv.maxCUSize, ctuY*pcv.maxCUSize), CH_L, TREE_D); |
493 | 0 | int mergedCTUPos; |
494 | 0 | int numValidMergeCandidates = 0; |
495 | |
|
496 | 0 | for(int mergeType=0; mergeType< NUM_SAO_MERGE_TYPES; mergeType++) |
497 | 0 | { |
498 | 0 | SAOBlkParam* mergeCandidate = NULL; |
499 | |
|
500 | 0 | switch(mergeType) |
501 | 0 | { |
502 | 0 | case SAO_MERGE_ABOVE: |
503 | 0 | { |
504 | 0 | if(ctuY > 0) |
505 | 0 | { |
506 | 0 | mergedCTUPos = ctuRsAddr- pcv.widthInCtus; |
507 | 0 | if(cs.getCURestricted(Position(ctuX*pcv.maxCUSize, (ctuY-1)*pcv.maxCUSize), cu, cu.chType)) |
508 | 0 | { |
509 | 0 | mergeCandidate = &(blkParams[mergedCTUPos]); |
510 | 0 | } |
511 | 0 | } |
512 | 0 | } |
513 | 0 | break; |
514 | 0 | case SAO_MERGE_LEFT: |
515 | 0 | { |
516 | 0 | if(ctuX > 0) |
517 | 0 | { |
518 | 0 | mergedCTUPos = ctuRsAddr- 1; |
519 | 0 | if(cs.getCURestricted(Position((ctuX-1)*pcv.maxCUSize, ctuY*pcv.maxCUSize), cu, cu.chType)) |
520 | 0 | { |
521 | 0 | mergeCandidate = &(blkParams[mergedCTUPos]); |
522 | 0 | } |
523 | 0 | } |
524 | 0 | } |
525 | 0 | break; |
526 | 0 | default: |
527 | 0 | { |
528 | 0 | THROW("not a supported merge type"); |
529 | 0 | } |
530 | 0 | } |
531 | | |
532 | 0 | mergeList[mergeType]=mergeCandidate; |
533 | 0 | if (mergeCandidate != NULL) |
534 | 0 | { |
535 | 0 | numValidMergeCandidates++; |
536 | 0 | } |
537 | 0 | } |
538 | | |
539 | 0 | return numValidMergeCandidates; |
540 | 0 | } |
541 | | |
542 | | |
543 | | void SampleAdaptiveOffset::reconstructBlkSAOParam(SAOBlkParam& recParam, SAOBlkParam* mergeList[NUM_SAO_MERGE_TYPES]) |
544 | 0 | { |
545 | 0 | const int numberOfComponents = m_numberOfComponents; |
546 | 0 | for(int compIdx = 0; compIdx < numberOfComponents; compIdx++) |
547 | 0 | { |
548 | 0 | const ComponentID component = ComponentID(compIdx); |
549 | 0 | SAOOffset& offsetParam = recParam[component]; |
550 | |
|
551 | 0 | if(offsetParam.modeIdc == SAO_MODE_OFF) |
552 | 0 | { |
553 | 0 | continue; |
554 | 0 | } |
555 | | |
556 | 0 | switch(offsetParam.modeIdc) |
557 | 0 | { |
558 | 0 | case SAO_MODE_NEW: |
559 | 0 | { |
560 | 0 | invertQuantOffsets(component, offsetParam.typeIdc, offsetParam.typeAuxInfo, offsetParam.offset, offsetParam.offset); |
561 | 0 | } |
562 | 0 | break; |
563 | 0 | case SAO_MODE_MERGE: |
564 | 0 | { |
565 | 0 | SAOBlkParam* mergeTarget = mergeList[offsetParam.typeIdc]; |
566 | 0 | CHECK(mergeTarget == NULL, "Merge target does not exist"); |
567 | |
|
568 | 0 | offsetParam = (*mergeTarget)[component]; |
569 | 0 | } |
570 | 0 | break; |
571 | 0 | default: |
572 | 0 | { |
573 | 0 | THROW("Not a supported mode"); |
574 | 0 | } |
575 | 0 | } |
576 | 0 | } |
577 | 0 | } |
578 | | |
579 | | void SampleAdaptiveOffset::xReconstructBlkSAOParams(CodingStructure& cs, SAOBlkParam* saoBlkParams) |
580 | 0 | { |
581 | 0 | for(uint32_t compIdx = 0; compIdx < MAX_NUM_COMP; compIdx++) |
582 | 0 | { |
583 | 0 | m_picSAOEnabled[compIdx] = false; |
584 | 0 | } |
585 | |
|
586 | 0 | const uint32_t numberOfComponents = getNumberValidComponents(cs.pcv->chrFormat); |
587 | |
|
588 | 0 | for(int ctuRsAddr=0; ctuRsAddr< cs.pcv->sizeInCtus; ctuRsAddr++) |
589 | 0 | { |
590 | 0 | SAOBlkParam* mergeList[NUM_SAO_MERGE_TYPES] = { NULL }; |
591 | 0 | getMergeList(cs, ctuRsAddr, saoBlkParams, mergeList); |
592 | |
|
593 | 0 | reconstructBlkSAOParam(saoBlkParams[ctuRsAddr], mergeList); |
594 | |
|
595 | 0 | for(uint32_t compIdx = 0; compIdx < numberOfComponents; compIdx++) |
596 | 0 | { |
597 | 0 | if(saoBlkParams[ctuRsAddr][compIdx].modeIdc != SAO_MODE_OFF) |
598 | 0 | { |
599 | 0 | m_picSAOEnabled[compIdx] = true; |
600 | 0 | } |
601 | 0 | } |
602 | 0 | } |
603 | 0 | } |
604 | | |
605 | | void SampleAdaptiveOffset::offsetCTU( const UnitArea& area, const CPelUnitBuf& src, PelUnitBuf& res, SAOBlkParam& saoblkParam, CodingStructure& cs) |
606 | 0 | { |
607 | 0 | const uint32_t numberOfComponents = getNumberValidComponents( area.chromaFormat ); |
608 | 0 | bool bAllOff=true; |
609 | 0 | for( uint32_t compIdx = 0; compIdx < numberOfComponents; compIdx++) |
610 | 0 | { |
611 | 0 | if (saoblkParam[compIdx].modeIdc != SAO_MODE_OFF) |
612 | 0 | { |
613 | 0 | bAllOff=false; |
614 | 0 | } |
615 | 0 | } |
616 | 0 | if (bAllOff) |
617 | 0 | { |
618 | 0 | return; |
619 | 0 | } |
620 | | |
621 | 0 | uint8_t availMask; |
622 | | //block boundary availability |
623 | 0 | deriveLoopFilterBoundaryAvailibility(cs, area.Y(), availMask); |
624 | |
|
625 | 0 | const size_t lineBufferSize = area.Y().width + 1; |
626 | 0 | if (m_signLineBuf1.size() < lineBufferSize) |
627 | 0 | { |
628 | 0 | m_signLineBuf1.resize(lineBufferSize); |
629 | 0 | m_signLineBuf2.resize(lineBufferSize); |
630 | 0 | } |
631 | |
|
632 | 0 | for(int compIdx = 0; compIdx < numberOfComponents; compIdx++) |
633 | 0 | { |
634 | 0 | const ComponentID compID = ComponentID(compIdx); |
635 | 0 | const CompArea& compArea = area.block(compID); |
636 | 0 | SAOOffset& ctbOffset = saoblkParam[compIdx]; |
637 | |
|
638 | 0 | if(ctbOffset.modeIdc != SAO_MODE_OFF) |
639 | 0 | { |
640 | 0 | int srcStride = src.get(compID).stride; |
641 | 0 | const Pel* srcBlk = src.get(compID).bufAt(compArea); |
642 | 0 | int resStride = res.get(compID).stride; |
643 | 0 | Pel* resBlk = res.get(compID).bufAt(compArea); |
644 | |
|
645 | 0 | offsetBlock( cs.sps->bitDepths[toChannelType(compID)], |
646 | 0 | cs.slice->clpRngs[compID], |
647 | 0 | ctbOffset.typeIdc, ctbOffset.offset, ctbOffset.typeAuxInfo |
648 | 0 | , srcBlk, resBlk, srcStride, resStride, compArea.width, compArea.height, availMask |
649 | 0 | , m_signLineBuf1, m_signLineBuf2 ); |
650 | 0 | } |
651 | 0 | } //compIdx |
652 | 0 | } |
653 | | |
654 | | void SampleAdaptiveOffset::deriveLoopFilterBoundaryAvailibility(CodingStructure& cs, const Position& pos, uint8_t& availMask ) const |
655 | 0 | { |
656 | 0 | const int cuSize = cs.pcv->maxCUSize; |
657 | 0 | CodingUnit *cuLeft, *cuRight, *cuAbove, *cuBelow, *cuAboveLeft, *cuAboveRight, *cuBelowLeft, *cuBelowRight; |
658 | 0 | const CodingUnit *cuCurr = cs.getCU( pos, CH_L, TREE_D ); |
659 | |
|
660 | 0 | if (!cs.pps->getSubPicFromCU(*cuCurr).loopFilterAcrossSubPicEnabled) |
661 | 0 | { |
662 | 0 | THROW("no support"); |
663 | 0 | } |
664 | | |
665 | 0 | if( !cs.pps->loopFilterAcrossSlicesEnabled || !cs.pps->loopFilterAcrossTilesEnabled ) |
666 | 0 | { |
667 | 0 | const int ctuX = pos.x >> cs.pcv->maxCUSizeLog2; |
668 | 0 | const int ctuY = pos.y >> cs.pcv->maxCUSizeLog2; |
669 | 0 | const PPS* pps = cs.slice->pps; |
670 | 0 | const int Xmax = pps->pcv->widthInCtus - 1; |
671 | 0 | const int Ymax = pps->pcv->heightInCtus - 1; |
672 | 0 | cuLeft = ctuX > 0 && pps->canFilterCtuBdry( ctuX, ctuY, -1, 0 ) ? cs.getCU( pos.offset( -cuSize, 0 ), CH_L, TREE_D ): nullptr; |
673 | 0 | cuRight = ctuX < Xmax && pps->canFilterCtuBdry( ctuX, ctuY, 1, 0 ) ? cs.getCU( pos.offset( cuSize , 0 ), CH_L, TREE_D ): nullptr; |
674 | 0 | cuAbove = ctuY > 0 && pps->canFilterCtuBdry( ctuX, ctuY, 0,-1 ) ? cs.getCU( pos.offset( 0 , -cuSize ), CH_L, TREE_D ): nullptr; |
675 | 0 | cuBelow = ctuY < Ymax && pps->canFilterCtuBdry( ctuX, ctuY, 0, 1 ) ? cs.getCU( pos.offset( 0 , cuSize ), CH_L, TREE_D ): nullptr; |
676 | 0 | cuAboveLeft = ctuX > 0 && ctuY > 0 && pps->canFilterCtuBdry( ctuX, ctuY, -1,-1 ) ? cs.getCU( pos.offset( -cuSize, -cuSize ), CH_L, TREE_D ): nullptr; |
677 | 0 | cuAboveRight = ctuX < Xmax && ctuY > 0 && pps->canFilterCtuBdry( ctuX, ctuY, 1,-1 ) ? cs.getCU( pos.offset( cuSize , -cuSize ), CH_L, TREE_D ): nullptr; |
678 | 0 | cuBelowLeft = ctuX > 0 && ctuY < Ymax && pps->canFilterCtuBdry( ctuX, ctuY, -1, 1 ) ? cs.getCU( pos.offset( -cuSize, cuSize ), CH_L, TREE_D ): nullptr; |
679 | 0 | cuBelowRight = ctuX < Xmax && ctuY < Ymax && pps->canFilterCtuBdry( ctuX, ctuY, 1, 1 ) ? cs.getCU( pos.offset( cuSize , cuSize ), CH_L, TREE_D ): nullptr; |
680 | 0 | } |
681 | 0 | else |
682 | 0 | { |
683 | 0 | cuLeft = cs.getCU( pos.offset( -cuSize, 0 ), CH_L, TREE_D ); |
684 | 0 | cuRight = cs.getCU( pos.offset( cuSize, 0 ), CH_L, TREE_D ); |
685 | 0 | cuAbove = cs.getCU( pos.offset( 0, -cuSize ), CH_L, TREE_D ); |
686 | 0 | cuBelow = cs.getCU( pos.offset( 0, cuSize ), CH_L, TREE_D ); |
687 | 0 | cuAboveLeft = cs.getCU( pos.offset( -cuSize, -cuSize ), CH_L, TREE_D ); |
688 | 0 | cuAboveRight = cs.getCU( pos.offset( cuSize, -cuSize ), CH_L, TREE_D ); |
689 | 0 | cuBelowLeft = cs.getCU( pos.offset( -cuSize, cuSize ), CH_L, TREE_D ); |
690 | 0 | cuBelowRight = cs.getCU( pos.offset( cuSize, cuSize ), CH_L, TREE_D ); |
691 | 0 | } |
692 | 0 | availMask = 0; |
693 | | |
694 | | // check cross slice flags |
695 | 0 | if( cs.pps->loopFilterAcrossSlicesEnabled ) |
696 | 0 | { |
697 | 0 | availMask |= (cuLeft != NULL) ? LeftAvail : 0; |
698 | 0 | availMask |= (cuAbove != NULL) ? AboveAvail : 0; |
699 | 0 | availMask |= (cuRight != NULL) ? RightAvail : 0; |
700 | 0 | availMask |= (cuBelow != NULL) ? BelowAvail : 0; |
701 | 0 | availMask |= (cuAboveLeft != NULL) ? AboveLeftAvail : 0; |
702 | 0 | availMask |= (cuBelowRight != NULL) ? BelowRightAvail : 0; |
703 | 0 | availMask |= (cuAboveRight != NULL) ? AboveRightAvail : 0; |
704 | 0 | availMask |= (cuBelowLeft != NULL) ? BelowLeftAvail : 0; |
705 | 0 | } |
706 | 0 | else |
707 | 0 | { |
708 | 0 | availMask |= ((cuLeft != NULL) && CU::isSameSlice(*cuCurr, *cuLeft) ) ? LeftAvail : 0; |
709 | 0 | availMask |= ((cuAbove != NULL) && CU::isSameSlice(*cuCurr, *cuAbove) ) ? AboveAvail : 0; |
710 | 0 | availMask |= ((cuRight != NULL) && CU::isSameSlice(*cuCurr, *cuRight)) ? RightAvail : 0; |
711 | 0 | availMask |= ((cuBelow != NULL) && CU::isSameSlice(*cuCurr, *cuBelow) ) ? BelowAvail : 0; |
712 | 0 | availMask |= ((cuAboveLeft != NULL) && CU::isSameSlice(*cuCurr, *cuAboveLeft)) ? AboveLeftAvail : 0; |
713 | 0 | availMask |= ((cuBelowRight != NULL) && CU::isSameSlice(*cuCurr, *cuBelowRight) ) ? BelowRightAvail : 0; |
714 | 0 | availMask |= ((cuAboveRight != NULL) && CU::isSameSlice(*cuCurr, *cuAboveRight) ) ? AboveRightAvail : 0; |
715 | 0 | availMask |= ( (cuBelowLeft != NULL) && CU::isSameSlice(*cuCurr, *cuBelowLeft) ) ? BelowLeftAvail : 0; |
716 | 0 | } |
717 | | |
718 | | // check cross tile flags |
719 | 0 | if (!cs.pps->loopFilterAcrossTilesEnabled) |
720 | 0 | { |
721 | 0 | uint8_t availMaskTile = 0; |
722 | 0 | availMaskTile |= (availMask&LeftAvail && CU::isSameTile(*cuCurr, *cuLeft)) ? LeftAvail : 0; |
723 | 0 | availMaskTile |= (availMask&AboveAvail && CU::isSameTile(*cuCurr, *cuAbove)) ? AboveAvail : 0; |
724 | 0 | availMaskTile |= (availMask&RightAvail && CU::isSameTile(*cuCurr, *cuRight)) ? RightAvail : 0; |
725 | 0 | availMaskTile |= (availMask&BelowAvail && CU::isSameTile(*cuCurr, *cuBelow)) ? BelowAvail : 0; |
726 | 0 | availMaskTile |= (availMask&AboveLeftAvail && CU::isSameTile(*cuCurr, *cuAboveLeft)) ? AboveLeftAvail : 0; |
727 | 0 | availMaskTile |= (availMask&AboveRightAvail &&CU::isSameTile(*cuCurr, *cuAboveRight)) ? AboveRightAvail : 0; |
728 | 0 | availMaskTile |= (availMask&BelowLeftAvail && CU::isSameTile(*cuCurr, *cuBelowLeft)) ? BelowLeftAvail : 0; |
729 | 0 | availMaskTile |= (availMask&BelowRightAvail && CU::isSameTile(*cuCurr, *cuBelowRight)) ? BelowRightAvail : 0; |
730 | 0 | availMask = availMaskTile; |
731 | 0 | } |
732 | 0 | } |
733 | | |
734 | | } // namespace vvenc |
735 | | |
736 | | //! \} |
737 | | |