1 /*
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
3 *
4 * This code is free software; you can redistribute it and/or modify it
5 * under the terms of the GNU General Public License version 2 only, as
6 * published by the Free Software Foundation. Oracle designates this
7 * particular file as subject to the "Classpath" exception as provided
8 * by Oracle in the LICENSE file that accompanied this code.
9 *
10 * This code is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
13 * version 2 for more details (a copy is included in the LICENSE file that
14 * accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License version
17 * 2 along with this work; if not, write to the Free Software Foundation,
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
19 *
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
21 * or visit www.oracle.com if you need additional information or have any
22 * questions.
23 */
24
25 // This file is available under and governed by the GNU General Public
26 // License version 2 only, as published by the Free Software Foundation.
27 // However, the following notice accompanied the original version of this
28 // file:
29 //
30 //---------------------------------------------------------------------------------
31 //
32 // Little Color Management System
33 // Copyright (c) 1998-2013 Marti Maria Saguer
34 //
35 // Permission is hereby granted, free of charge, to any person obtaining
36 // a copy of this software and associated documentation files (the "Software"),
37 // to deal in the Software without restriction, including without limitation
38 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
39 // and/or sell copies of the Software, and to permit persons to whom the Software
40 // is furnished to do so, subject to the following conditions:
41 //
42 // The above copyright notice and this permission notice shall be included in
43 // all copies or substantial portions of the Software.
44 //
45 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
46 // EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
47 // THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
48 // NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
49 // LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
50 // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
51 // WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
52 //
53 //---------------------------------------------------------------------------------
54 //
55 #include "lcms2_internal.h"
56
57 // Tone curves are powerful constructs that can contain curves specified in diverse ways.
58 // The curve is stored in segments, where each segment can be sampled or specified by parameters.
59 // a 16.bit simplification of the *whole* curve is kept for optimization purposes. For float operation,
60 // each segment is evaluated separately. Plug-ins may be used to define new parametric schemes,
61 // each plug-in may define up to MAX_TYPES_IN_LCMS_PLUGIN functions types. For defining a function,
62 // the plug-in should provide the type id, how many parameters each type has, and a pointer to
63 // a procedure that evaluates the function. In the case of reverse evaluation, the evaluator will
64 // be called with the type id as a negative value, and a sampled version of the reversed curve
65 // will be built.
66
67 // ----------------------------------------------------------------- Implementation
68 // Maxim number of nodes
69 #define MAX_NODES_IN_CURVE 4097
70 #define MINUS_INF (-1E22F)
71 #define PLUS_INF (+1E22F)
72
73 // The list of supported parametric curves
74 typedef struct _cmsParametricCurvesCollection_st {
75
76 cmsUInt32Number nFunctions; // Number of supported functions in this chunk
77 cmsInt32Number FunctionTypes[MAX_TYPES_IN_LCMS_PLUGIN]; // The identification types
78 cmsUInt32Number ParameterCount[MAX_TYPES_IN_LCMS_PLUGIN]; // Number of parameters for each function
79
80 cmsParametricCurveEvaluator Evaluator; // The evaluator
81
82 struct _cmsParametricCurvesCollection_st* Next; // Next in list
83
84 } _cmsParametricCurvesCollection;
85
86 // This is the default (built-in) evaluator
87 static cmsFloat64Number DefaultEvalParametricFn(cmsInt32Number Type, const cmsFloat64Number Params[], cmsFloat64Number R);
88
89 // The built-in list
90 static _cmsParametricCurvesCollection DefaultCurves = {
91 9, // # of curve types
92 { 1, 2, 3, 4, 5, 6, 7, 8, 108 }, // Parametric curve ID
93 { 1, 3, 4, 5, 7, 4, 5, 5, 1 }, // Parameters by type
94 DefaultEvalParametricFn, // Evaluator
95 NULL // Next in chain
96 };
97
98 // Duplicates the zone of memory used by the plug-in in the new context
99 static
100 void DupPluginCurvesList(struct _cmsContext_struct* ctx,
101 const struct _cmsContext_struct* src)
102 {
103 _cmsCurvesPluginChunkType newHead = { NULL };
104 _cmsParametricCurvesCollection* entry;
105 _cmsParametricCurvesCollection* Anterior = NULL;
106 _cmsCurvesPluginChunkType* head = (_cmsCurvesPluginChunkType*) src->chunks[CurvesPlugin];
107
108 _cmsAssert(head != NULL);
109
110 // Walk the list copying all nodes
111 for (entry = head->ParametricCurves;
112 entry != NULL;
113 entry = entry ->Next) {
114
115 _cmsParametricCurvesCollection *newEntry = ( _cmsParametricCurvesCollection *) _cmsSubAllocDup(ctx ->MemPool, entry, sizeof(_cmsParametricCurvesCollection));
116
117 if (newEntry == NULL)
118 return;
119
120 // We want to keep the linked list order, so this is a little bit tricky
121 newEntry -> Next = NULL;
122 if (Anterior)
123 Anterior -> Next = newEntry;
124
125 Anterior = newEntry;
126
127 if (newHead.ParametricCurves == NULL)
128 newHead.ParametricCurves = newEntry;
129 }
130
131 ctx ->chunks[CurvesPlugin] = _cmsSubAllocDup(ctx->MemPool, &newHead, sizeof(_cmsCurvesPluginChunkType));
132 }
133
134 // The allocator have to follow the chain
135 void _cmsAllocCurvesPluginChunk(struct _cmsContext_struct* ctx,
136 const struct _cmsContext_struct* src)
137 {
138 _cmsAssert(ctx != NULL);
139
140 if (src != NULL) {
141
142 // Copy all linked list
143 DupPluginCurvesList(ctx, src);
144 }
145 else {
146 static _cmsCurvesPluginChunkType CurvesPluginChunk = { NULL };
147 ctx ->chunks[CurvesPlugin] = _cmsSubAllocDup(ctx ->MemPool, &CurvesPluginChunk, sizeof(_cmsCurvesPluginChunkType));
148 }
149 }
150
151
152 // The linked list head
153 _cmsCurvesPluginChunkType _cmsCurvesPluginChunk = { NULL };
154
155 // As a way to install new parametric curves
156 cmsBool _cmsRegisterParametricCurvesPlugin(cmsContext ContextID, cmsPluginBase* Data)
157 {
158 _cmsCurvesPluginChunkType* ctx = ( _cmsCurvesPluginChunkType*) _cmsContextGetClientChunk(ContextID, CurvesPlugin);
159 cmsPluginParametricCurves* Plugin = (cmsPluginParametricCurves*) Data;
160 _cmsParametricCurvesCollection* fl;
161
162 if (Data == NULL) {
163
164 ctx -> ParametricCurves = NULL;
165 return TRUE;
166 }
167
168 fl = (_cmsParametricCurvesCollection*) _cmsPluginMalloc(ContextID, sizeof(_cmsParametricCurvesCollection));
169 if (fl == NULL) return FALSE;
170
171 // Copy the parameters
172 fl ->Evaluator = Plugin ->Evaluator;
173 fl ->nFunctions = Plugin ->nFunctions;
174
175 // Make sure no mem overwrites
176 if (fl ->nFunctions > MAX_TYPES_IN_LCMS_PLUGIN)
177 fl ->nFunctions = MAX_TYPES_IN_LCMS_PLUGIN;
178
179 // Copy the data
180 memmove(fl->FunctionTypes, Plugin ->FunctionTypes, fl->nFunctions * sizeof(cmsUInt32Number));
181 memmove(fl->ParameterCount, Plugin ->ParameterCount, fl->nFunctions * sizeof(cmsUInt32Number));
182
183 // Keep linked list
184 fl ->Next = ctx->ParametricCurves;
185 ctx->ParametricCurves = fl;
186
187 // All is ok
188 return TRUE;
189 }
190
191
192 // Search in type list, return position or -1 if not found
193 static
194 int IsInSet(int Type, _cmsParametricCurvesCollection* c)
195 {
196 int i;
197
198 for (i=0; i < (int) c ->nFunctions; i++)
199 if (abs(Type) == c ->FunctionTypes[i]) return i;
200
201 return -1;
202 }
203
204
205 // Search for the collection which contains a specific type
206 static
207 _cmsParametricCurvesCollection *GetParametricCurveByType(cmsContext ContextID, int Type, int* index)
208 {
209 _cmsParametricCurvesCollection* c;
210 int Position;
211 _cmsCurvesPluginChunkType* ctx = ( _cmsCurvesPluginChunkType*) _cmsContextGetClientChunk(ContextID, CurvesPlugin);
212
213 for (c = ctx->ParametricCurves; c != NULL; c = c ->Next) {
214
215 Position = IsInSet(Type, c);
216
217 if (Position != -1) {
218 if (index != NULL)
219 *index = Position;
220 return c;
221 }
222 }
223 // If none found, revert for defaults
224 for (c = &DefaultCurves; c != NULL; c = c ->Next) {
225
226 Position = IsInSet(Type, c);
227
228 if (Position != -1) {
229 if (index != NULL)
230 *index = Position;
231 return c;
232 }
233 }
234
235 return NULL;
236 }
237
238 // Low level allocate, which takes care of memory details. nEntries may be zero, and in this case
239 // no optimation curve is computed. nSegments may also be zero in the inverse case, where only the
240 // optimization curve is given. Both features simultaneously is an error
241 static
242 cmsToneCurve* AllocateToneCurveStruct(cmsContext ContextID, cmsUInt32Number nEntries,
243 cmsUInt32Number nSegments, const cmsCurveSegment* Segments,
244 const cmsUInt16Number* Values)
245 {
246 cmsToneCurve* p;
247 cmsUInt32Number i;
248
249 // We allow huge tables, which are then restricted for smoothing operations
250 if (nEntries > 65530) {
251 cmsSignalError(ContextID, cmsERROR_RANGE, "Couldn't create tone curve of more than 65530 entries");
252 return NULL;
253 }
254
255 if (nEntries == 0 && nSegments == 0) {
256 cmsSignalError(ContextID, cmsERROR_RANGE, "Couldn't create tone curve with zero segments and no table");
257 return NULL;
258 }
259
260 // Allocate all required pointers, etc.
261 p = (cmsToneCurve*) _cmsMallocZero(ContextID, sizeof(cmsToneCurve));
262 if (!p) return NULL;
263
264 // In this case, there are no segments
265 if (nSegments == 0) {
266 p ->Segments = NULL;
267 p ->Evals = NULL;
268 }
269 else {
270 p ->Segments = (cmsCurveSegment*) _cmsCalloc(ContextID, nSegments, sizeof(cmsCurveSegment));
271 if (p ->Segments == NULL) goto Error;
272
273 p ->Evals = (cmsParametricCurveEvaluator*) _cmsCalloc(ContextID, nSegments, sizeof(cmsParametricCurveEvaluator));
274 if (p ->Evals == NULL) goto Error;
275 }
276
277 p -> nSegments = nSegments;
278
279 // This 16-bit table contains a limited precision representation of the whole curve and is kept for
280 // increasing xput on certain operations.
281 if (nEntries == 0) {
282 p ->Table16 = NULL;
283 }
284 else {
285 p ->Table16 = (cmsUInt16Number*) _cmsCalloc(ContextID, nEntries, sizeof(cmsUInt16Number));
286 if (p ->Table16 == NULL) goto Error;
287 }
288
289 p -> nEntries = nEntries;
290
291 // Initialize members if requested
292 if (Values != NULL && (nEntries > 0)) {
293
294 for (i=0; i < nEntries; i++)
295 p ->Table16[i] = Values[i];
296 }
297
298 // Initialize the segments stuff. The evaluator for each segment is located and a pointer to it
299 // is placed in advance to maximize performance.
300 if (Segments != NULL && (nSegments > 0)) {
301
302 _cmsParametricCurvesCollection *c;
303
304 p ->SegInterp = (cmsInterpParams**) _cmsCalloc(ContextID, nSegments, sizeof(cmsInterpParams*));
305 if (p ->SegInterp == NULL) goto Error;
306
307 for (i=0; i < nSegments; i++) {
308
309 // Type 0 is a special marker for table-based curves
310 if (Segments[i].Type == 0)
311 p ->SegInterp[i] = _cmsComputeInterpParams(ContextID, Segments[i].nGridPoints, 1, 1, NULL, CMS_LERP_FLAGS_FLOAT);
312
313 memmove(&p ->Segments[i], &Segments[i], sizeof(cmsCurveSegment));
314
315 if (Segments[i].Type == 0 && Segments[i].SampledPoints != NULL)
316 p ->Segments[i].SampledPoints = (cmsFloat32Number*) _cmsDupMem(ContextID, Segments[i].SampledPoints, sizeof(cmsFloat32Number) * Segments[i].nGridPoints);
317 else
318 p ->Segments[i].SampledPoints = NULL;
319
320
321 c = GetParametricCurveByType(ContextID, Segments[i].Type, NULL);
322 if (c != NULL)
323 p ->Evals[i] = c ->Evaluator;
324 }
325 }
326
327 p ->InterpParams = _cmsComputeInterpParams(ContextID, p ->nEntries, 1, 1, p->Table16, CMS_LERP_FLAGS_16BITS);
328 if (p->InterpParams != NULL)
329 return p;
330
331 Error:
332 if (p -> Segments) _cmsFree(ContextID, p ->Segments);
333 if (p -> Evals) _cmsFree(ContextID, p -> Evals);
334 if (p ->Table16) _cmsFree(ContextID, p ->Table16);
335 _cmsFree(ContextID, p);
336 return NULL;
337 }
338
339
340 // Parametric Fn using floating point
341 static
342 cmsFloat64Number DefaultEvalParametricFn(cmsInt32Number Type, const cmsFloat64Number Params[], cmsFloat64Number R)
343 {
344 cmsFloat64Number e, Val, disc;
345
346 switch (Type) {
347
348 // X = Y ^ Gamma
349 case 1:
350 if (R < 0) {
351
352 if (fabs(Params[0] - 1.0) < MATRIX_DET_TOLERANCE)
353 Val = R;
354 else
355 Val = 0;
356 }
357 else
358 Val = pow(R, Params[0]);
359 break;
360
361 // Type 1 Reversed: X = Y ^1/gamma
362 case -1:
363 if (R < 0) {
364
365 if (fabs(Params[0] - 1.0) < MATRIX_DET_TOLERANCE)
366 Val = R;
367 else
368 Val = 0;
369 }
370 else
371 {
372 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE)
373 Val = PLUS_INF;
374 else
375 Val = pow(R, 1 / Params[0]);
376 }
377 break;
378
379 // CIE 122-1966
380 // Y = (aX + b)^Gamma | X >= -b/a
381 // Y = 0 | else
382 case 2:
383 {
384
385 if (fabs(Params[1]) < MATRIX_DET_TOLERANCE)
386 {
387 Val = 0;
388 }
389 else
390 {
391 disc = -Params[2] / Params[1];
392
393 if (R >= disc) {
394
395 e = Params[1] * R + Params[2];
396
397 if (e > 0)
398 Val = pow(e, Params[0]);
399 else
400 Val = 0;
401 }
402 else
403 Val = 0;
404 }
405 }
406 break;
407
408 // Type 2 Reversed
409 // X = (Y ^1/g - b) / a
410 case -2:
411 {
412 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE ||
413 fabs(Params[1]) < MATRIX_DET_TOLERANCE)
414 {
415 Val = 0;
416 }
417 else
418 {
419 if (R < 0)
420 Val = 0;
421 else
422 Val = (pow(R, 1.0 / Params[0]) - Params[2]) / Params[1];
423
424 if (Val < 0)
425 Val = 0;
426 }
427 }
428 break;
429
430
431 // IEC 61966-3
432 // Y = (aX + b)^Gamma | X <= -b/a
433 // Y = c | else
434 case 3:
435 {
436 if (fabs(Params[1]) < MATRIX_DET_TOLERANCE)
437 {
438 Val = 0;
439 }
440 else
441 {
442 disc = -Params[2] / Params[1];
443 if (disc < 0)
444 disc = 0;
445
446 if (R >= disc) {
447
448 e = Params[1] * R + Params[2];
449
450 if (e > 0)
451 Val = pow(e, Params[0]) + Params[3];
452 else
453 Val = 0;
454 }
455 else
456 Val = Params[3];
457 }
458 }
459 break;
460
461
462 // Type 3 reversed
463 // X=((Y-c)^1/g - b)/a | (Y>=c)
464 // X=-b/a | (Y<c)
465 case -3:
466 {
467 if (fabs(Params[1]) < MATRIX_DET_TOLERANCE)
468 {
469 Val = 0;
470 }
471 else
472 {
473 if (R >= Params[3]) {
474
475 e = R - Params[3];
476
477 if (e > 0)
478 Val = (pow(e, 1 / Params[0]) - Params[2]) / Params[1];
479 else
480 Val = 0;
481 }
482 else {
483 Val = -Params[2] / Params[1];
484 }
485 }
486 }
487 break;
488
489
490 // IEC 61966-2.1 (sRGB)
491 // Y = (aX + b)^Gamma | X >= d
492 // Y = cX | X < d
493 case 4:
494 if (R >= Params[4]) {
495
496 e = Params[1]*R + Params[2];
497
498 if (e > 0)
499 Val = pow(e, Params[0]);
500 else
501 Val = 0;
502 }
503 else
504 Val = R * Params[3];
505 break;
506
507 // Type 4 reversed
508 // X=((Y^1/g-b)/a) | Y >= (ad+b)^g
509 // X=Y/c | Y< (ad+b)^g
510 case -4:
511 {
512 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE ||
513 fabs(Params[1]) < MATRIX_DET_TOLERANCE ||
514 fabs(Params[3]) < MATRIX_DET_TOLERANCE)
515 {
516 Val = 0;
517 }
518 else
519 {
520 e = Params[1] * Params[4] + Params[2];
521 if (e < 0)
522 disc = 0;
523 else
524 disc = pow(e, Params[0]);
525
526 if (R >= disc) {
527
528 Val = (pow(R, 1.0 / Params[0]) - Params[2]) / Params[1];
529 }
530 else {
531 Val = R / Params[3];
532 }
533 }
534 }
535 break;
536
537
538 // Y = (aX + b)^Gamma + e | X >= d
539 // Y = cX + f | X < d
540 case 5:
541 if (R >= Params[4]) {
542
543 e = Params[1]*R + Params[2];
544
545 if (e > 0)
546 Val = pow(e, Params[0]) + Params[5];
547 else
548 Val = Params[5];
549 }
550 else
551 Val = R*Params[3] + Params[6];
552 break;
553
554
555 // Reversed type 5
556 // X=((Y-e)1/g-b)/a | Y >=(ad+b)^g+e), cd+f
557 // X=(Y-f)/c | else
558 case -5:
559 {
560 if (fabs(Params[1]) < MATRIX_DET_TOLERANCE ||
561 fabs(Params[3]) < MATRIX_DET_TOLERANCE)
562 {
563 Val = 0;
564 }
565 else
566 {
567 disc = Params[3] * Params[4] + Params[6];
568 if (R >= disc) {
569
570 e = R - Params[5];
571 if (e < 0)
572 Val = 0;
573 else
574 Val = (pow(e, 1.0 / Params[0]) - Params[2]) / Params[1];
575 }
576 else {
577 Val = (R - Params[6]) / Params[3];
578 }
579 }
580 }
581 break;
582
583
584 // Types 6,7,8 comes from segmented curves as described in ICCSpecRevision_02_11_06_Float.pdf
585 // Type 6 is basically identical to type 5 without d
586
587 // Y = (a * X + b) ^ Gamma + c
588 case 6:
589 e = Params[1]*R + Params[2];
590
591 if (e < 0)
592 Val = Params[3];
593 else
594 Val = pow(e, Params[0]) + Params[3];
595 break;
596
597 // ((Y - c) ^1/Gamma - b) / a
598 case -6:
599 {
600 if (fabs(Params[1]) < MATRIX_DET_TOLERANCE)
601 {
602 Val = 0;
603 }
604 else
605 {
606 e = R - Params[3];
607 if (e < 0)
608 Val = 0;
609 else
610 Val = (pow(e, 1.0 / Params[0]) - Params[2]) / Params[1];
611 }
612 }
613 break;
614
615
616 // Y = a * log (b * X^Gamma + c) + d
617 case 7:
618
619 e = Params[2] * pow(R, Params[0]) + Params[3];
620 if (e <= 0)
621 Val = Params[4];
622 else
623 Val = Params[1]*log10(e) + Params[4];
624 break;
625
626 // (Y - d) / a = log(b * X ^Gamma + c)
627 // pow(10, (Y-d) / a) = b * X ^Gamma + c
628 // pow((pow(10, (Y-d) / a) - c) / b, 1/g) = X
629 case -7:
630 {
631 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE ||
632 fabs(Params[1]) < MATRIX_DET_TOLERANCE ||
633 fabs(Params[2]) < MATRIX_DET_TOLERANCE)
634 {
635 Val = 0;
636 }
637 else
638 {
639 Val = pow((pow(10.0, (R - Params[4]) / Params[1]) - Params[3]) / Params[2], 1.0 / Params[0]);
640 }
641 }
642 break;
643
644
645 //Y = a * b^(c*X+d) + e
646 case 8:
647 Val = (Params[0] * pow(Params[1], Params[2] * R + Params[3]) + Params[4]);
648 break;
649
650
651 // Y = (log((y-e) / a) / log(b) - d ) / c
652 // a=0, b=1, c=2, d=3, e=4,
653 case -8:
654
655 disc = R - Params[4];
656 if (disc < 0) Val = 0;
657 else
658 {
659 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE ||
660 fabs(Params[2]) < MATRIX_DET_TOLERANCE)
661 {
662 Val = 0;
663 }
664 else
665 {
666 Val = (log(disc / Params[0]) / log(Params[1]) - Params[3]) / Params[2];
667 }
668 }
669 break;
670
671 // S-Shaped: (1 - (1-x)^1/g)^1/g
672 case 108:
673 if (fabs(Params[0]) < MATRIX_DET_TOLERANCE)
674 Val = 0;
675 else
676 Val = pow(1.0 - pow(1 - R, 1/Params[0]), 1/Params[0]);
677 break;
678
679 // y = (1 - (1-x)^1/g)^1/g
680 // y^g = (1 - (1-x)^1/g)
681 // 1 - y^g = (1-x)^1/g
682 // (1 - y^g)^g = 1 - x
683 // 1 - (1 - y^g)^g
684 case -108:
685 Val = 1 - pow(1 - pow(R, Params[0]), Params[0]);
686 break;
687
688 default:
689 // Unsupported parametric curve. Should never reach here
690 return 0;
691 }
692
693 return Val;
694 }
695
696 // Evaluate a segmented function for a single value. Return -Inf if no valid segment found .
697 // If fn type is 0, perform an interpolation on the table
698 static
699 cmsFloat64Number EvalSegmentedFn(const cmsToneCurve *g, cmsFloat64Number R)
700 {
701 int i;
702 cmsFloat32Number Out32;
703 cmsFloat64Number Out;
704
705 for (i = (int) g->nSegments - 1; i >= 0; --i) {
706
707 // Check for domain
708 if ((R > g->Segments[i].x0) && (R <= g->Segments[i].x1)) {
709
710 // Type == 0 means segment is sampled
711 if (g->Segments[i].Type == 0) {
712
713 cmsFloat32Number R1 = (cmsFloat32Number)(R - g->Segments[i].x0) / (g->Segments[i].x1 - g->Segments[i].x0);
714
715 // Setup the table (TODO: clean that)
716 g->SegInterp[i]->Table = g->Segments[i].SampledPoints;
717
718 g->SegInterp[i]->Interpolation.LerpFloat(&R1, &Out32, g->SegInterp[i]);
719 Out = (cmsFloat64Number) Out32;
720
721 }
722 else {
723 Out = g->Evals[i](g->Segments[i].Type, g->Segments[i].Params, R);
724 }
725
726 if (isinf(Out))
727 return PLUS_INF;
728 else
729 {
730 if (isinf(-Out))
731 return MINUS_INF;
732 }
733
734 return Out;
735 }
736 }
737
738 return MINUS_INF;
739 }
740
741 // Access to estimated low-res table
742 cmsUInt32Number CMSEXPORT cmsGetToneCurveEstimatedTableEntries(const cmsToneCurve* t)
743 {
744 _cmsAssert(t != NULL);
745 return t ->nEntries;
746 }
747
748 const cmsUInt16Number* CMSEXPORT cmsGetToneCurveEstimatedTable(const cmsToneCurve* t)
749 {
750 _cmsAssert(t != NULL);
751 return t ->Table16;
752 }
753
754
755 // Create an empty gamma curve, by using tables. This specifies only the limited-precision part, and leaves the
756 // floating point description empty.
757 cmsToneCurve* CMSEXPORT cmsBuildTabulatedToneCurve16(cmsContext ContextID, cmsUInt32Number nEntries, const cmsUInt16Number Values[])
758 {
759 return AllocateToneCurveStruct(ContextID, nEntries, 0, NULL, Values);
760 }
761
762 static
763 cmsUInt32Number EntriesByGamma(cmsFloat64Number Gamma)
764 {
765 if (fabs(Gamma - 1.0) < 0.001) return 2;
766 return 4096;
767 }
768
769
770 // Create a segmented gamma, fill the table
771 cmsToneCurve* CMSEXPORT cmsBuildSegmentedToneCurve(cmsContext ContextID,
772 cmsUInt32Number nSegments, const cmsCurveSegment Segments[])
773 {
774 cmsUInt32Number i;
775 cmsFloat64Number R, Val;
776 cmsToneCurve* g;
777 cmsUInt32Number nGridPoints = 4096;
778
779 _cmsAssert(Segments != NULL);
780
781 // Optimizatin for identity curves.
782 if (nSegments == 1 && Segments[0].Type == 1) {
783
784 nGridPoints = EntriesByGamma(Segments[0].Params[0]);
785 }
786
787 g = AllocateToneCurveStruct(ContextID, nGridPoints, nSegments, Segments, NULL);
788 if (g == NULL) return NULL;
789
790 // Once we have the floating point version, we can approximate a 16 bit table of 4096 entries
791 // for performance reasons. This table would normally not be used except on 8/16 bits transforms.
792 for (i = 0; i < nGridPoints; i++) {
793
794 R = (cmsFloat64Number) i / (nGridPoints-1);
795
796 Val = EvalSegmentedFn(g, R);
797
798 // Round and saturate
799 g ->Table16[i] = _cmsQuickSaturateWord(Val * 65535.0);
800 }
801
802 return g;
803 }
804
805 // Use a segmented curve to store the floating point table
806 cmsToneCurve* CMSEXPORT cmsBuildTabulatedToneCurveFloat(cmsContext ContextID, cmsUInt32Number nEntries, const cmsFloat32Number values[])
807 {
808 cmsCurveSegment Seg[3];
809
810 // A segmented tone curve should have function segments in the first and last positions
811 // Initialize segmented curve part up to 0 to constant value = samples[0]
812 Seg[0].x0 = MINUS_INF;
813 Seg[0].x1 = 0;
814 Seg[0].Type = 6;
815
816 Seg[0].Params[0] = 1;
817 Seg[0].Params[1] = 0;
818 Seg[0].Params[2] = 0;
819 Seg[0].Params[3] = values[0];
820 Seg[0].Params[4] = 0;
821
822 // From zero to 1
823 Seg[1].x0 = 0;
824 Seg[1].x1 = 1.0;
825 Seg[1].Type = 0;
826
827 Seg[1].nGridPoints = nEntries;
828 Seg[1].SampledPoints = (cmsFloat32Number*) values;
829
830 // Final segment is constant = lastsample
831 Seg[2].x0 = 1.0;
832 Seg[2].x1 = PLUS_INF;
833 Seg[2].Type = 6;
834
835 Seg[2].Params[0] = 1;
836 Seg[2].Params[1] = 0;
837 Seg[2].Params[2] = 0;
838 Seg[2].Params[3] = values[nEntries-1];
839 Seg[2].Params[4] = 0;
840
841
842 return cmsBuildSegmentedToneCurve(ContextID, 3, Seg);
843 }
844
845 // Parametric curves
846 //
847 // Parameters goes as: Curve, a, b, c, d, e, f
848 // Type is the ICC type +1
849 // if type is negative, then the curve is analyticaly inverted
850 cmsToneCurve* CMSEXPORT cmsBuildParametricToneCurve(cmsContext ContextID, cmsInt32Number Type, const cmsFloat64Number Params[])
851 {
852 cmsCurveSegment Seg0;
853 int Pos = 0;
854 cmsUInt32Number size;
855 _cmsParametricCurvesCollection* c = GetParametricCurveByType(ContextID, Type, &Pos);
856
857 _cmsAssert(Params != NULL);
858
859 if (c == NULL) {
860 cmsSignalError(ContextID, cmsERROR_UNKNOWN_EXTENSION, "Invalid parametric curve type %d", Type);
861 return NULL;
862 }
863
864 memset(&Seg0, 0, sizeof(Seg0));
865
866 Seg0.x0 = MINUS_INF;
867 Seg0.x1 = PLUS_INF;
868 Seg0.Type = Type;
869
870 size = c->ParameterCount[Pos] * sizeof(cmsFloat64Number);
871 memmove(Seg0.Params, Params, size);
872
873 return cmsBuildSegmentedToneCurve(ContextID, 1, &Seg0);
874 }
875
876
877
878 // Build a gamma table based on gamma constant
879 cmsToneCurve* CMSEXPORT cmsBuildGamma(cmsContext ContextID, cmsFloat64Number Gamma)
880 {
881 return cmsBuildParametricToneCurve(ContextID, 1, &Gamma);
882 }
883
884
885 // Free all memory taken by the gamma curve
886 void CMSEXPORT cmsFreeToneCurve(cmsToneCurve* Curve)
887 {
888 cmsContext ContextID;
889
890 if (Curve == NULL) return;
891
892 ContextID = Curve ->InterpParams->ContextID;
893
894 _cmsFreeInterpParams(Curve ->InterpParams);
895
896 if (Curve -> Table16)
897 _cmsFree(ContextID, Curve ->Table16);
898
899 if (Curve ->Segments) {
900
901 cmsUInt32Number i;
902
903 for (i=0; i < Curve ->nSegments; i++) {
904
905 if (Curve ->Segments[i].SampledPoints) {
906 _cmsFree(ContextID, Curve ->Segments[i].SampledPoints);
907 }
908
909 if (Curve ->SegInterp[i] != 0)
910 _cmsFreeInterpParams(Curve->SegInterp[i]);
911 }
912
913 _cmsFree(ContextID, Curve ->Segments);
914 _cmsFree(ContextID, Curve ->SegInterp);
915 }
916
917 if (Curve -> Evals)
918 _cmsFree(ContextID, Curve -> Evals);
919
920 if (Curve) _cmsFree(ContextID, Curve);
921 }
922
923 // Utility function, free 3 gamma tables
924 void CMSEXPORT cmsFreeToneCurveTriple(cmsToneCurve* Curve[3])
925 {
926
927 _cmsAssert(Curve != NULL);
928
929 if (Curve[0] != NULL) cmsFreeToneCurve(Curve[0]);
930 if (Curve[1] != NULL) cmsFreeToneCurve(Curve[1]);
931 if (Curve[2] != NULL) cmsFreeToneCurve(Curve[2]);
932
933 Curve[0] = Curve[1] = Curve[2] = NULL;
934 }
935
936
937 // Duplicate a gamma table
938 cmsToneCurve* CMSEXPORT cmsDupToneCurve(const cmsToneCurve* In)
939 {
940 if (In == NULL) return NULL;
941
942 return AllocateToneCurveStruct(In ->InterpParams ->ContextID, In ->nEntries, In ->nSegments, In ->Segments, In ->Table16);
943 }
944
945 // Joins two curves for X and Y. Curves should be monotonic.
946 // We want to get
947 //
948 // y = Y^-1(X(t))
949 //
950 cmsToneCurve* CMSEXPORT cmsJoinToneCurve(cmsContext ContextID,
951 const cmsToneCurve* X,
952 const cmsToneCurve* Y, cmsUInt32Number nResultingPoints)
953 {
954 cmsToneCurve* out = NULL;
955 cmsToneCurve* Yreversed = NULL;
956 cmsFloat32Number t, x;
957 cmsFloat32Number* Res = NULL;
958 cmsUInt32Number i;
959
960
961 _cmsAssert(X != NULL);
962 _cmsAssert(Y != NULL);
963
964 Yreversed = cmsReverseToneCurveEx(nResultingPoints, Y);
965 if (Yreversed == NULL) goto Error;
966
967 Res = (cmsFloat32Number*) _cmsCalloc(ContextID, nResultingPoints, sizeof(cmsFloat32Number));
968 if (Res == NULL) goto Error;
969
970 //Iterate
971 for (i=0; i < nResultingPoints; i++) {
972
973 t = (cmsFloat32Number) i / (nResultingPoints-1);
974 x = cmsEvalToneCurveFloat(X, t);
975 Res[i] = cmsEvalToneCurveFloat(Yreversed, x);
976 }
977
978 // Allocate space for output
979 out = cmsBuildTabulatedToneCurveFloat(ContextID, nResultingPoints, Res);
980
981 Error:
982
983 if (Res != NULL) _cmsFree(ContextID, Res);
984 if (Yreversed != NULL) cmsFreeToneCurve(Yreversed);
985
986 return out;
987 }
988
989
990
991 // Get the surrounding nodes. This is tricky on non-monotonic tables
992 static
993 int GetInterval(cmsFloat64Number In, const cmsUInt16Number LutTable[], const struct _cms_interp_struc* p)
994 {
995 int i;
996 int y0, y1;
997
998 // A 1 point table is not allowed
999 if (p -> Domain[0] < 1) return -1;
1000
1001 // Let's see if ascending or descending.
1002 if (LutTable[0] < LutTable[p ->Domain[0]]) {
1003
1004 // Table is overall ascending
1005 for (i = (int) p->Domain[0] - 1; i >= 0; --i) {
1006
1007 y0 = LutTable[i];
1008 y1 = LutTable[i+1];
1009
1010 if (y0 <= y1) { // Increasing
1011 if (In >= y0 && In <= y1) return i;
1012 }
1013 else
1014 if (y1 < y0) { // Decreasing
1015 if (In >= y1 && In <= y0) return i;
1016 }
1017 }
1018 }
1019 else {
1020 // Table is overall descending
1021 for (i=0; i < (int) p -> Domain[0]; i++) {
1022
1023 y0 = LutTable[i];
1024 y1 = LutTable[i+1];
1025
1026 if (y0 <= y1) { // Increasing
1027 if (In >= y0 && In <= y1) return i;
1028 }
1029 else
1030 if (y1 < y0) { // Decreasing
1031 if (In >= y1 && In <= y0) return i;
1032 }
1033 }
1034 }
1035
1036 return -1;
1037 }
1038
1039 // Reverse a gamma table
1040 cmsToneCurve* CMSEXPORT cmsReverseToneCurveEx(cmsUInt32Number nResultSamples, const cmsToneCurve* InCurve)
1041 {
1042 cmsToneCurve *out;
1043 cmsFloat64Number a = 0, b = 0, y, x1, y1, x2, y2;
1044 int i, j;
1045 int Ascending;
1046
1047 _cmsAssert(InCurve != NULL);
1048
1049 // Try to reverse it analytically whatever possible
1050
1051 if (InCurve ->nSegments == 1 && InCurve ->Segments[0].Type > 0 &&
1052 /* InCurve -> Segments[0].Type <= 5 */
1053 GetParametricCurveByType(InCurve ->InterpParams->ContextID, InCurve ->Segments[0].Type, NULL) != NULL) {
1054
1055 return cmsBuildParametricToneCurve(InCurve ->InterpParams->ContextID,
1056 -(InCurve -> Segments[0].Type),
1057 InCurve -> Segments[0].Params);
1058 }
1059
1060 // Nope, reverse the table.
1061 out = cmsBuildTabulatedToneCurve16(InCurve ->InterpParams->ContextID, nResultSamples, NULL);
1062 if (out == NULL)
1063 return NULL;
1064
1065 // We want to know if this is an ascending or descending table
1066 Ascending = !cmsIsToneCurveDescending(InCurve);
1067
1068 // Iterate across Y axis
1069 for (i=0; i < (int) nResultSamples; i++) {
1070
1071 y = (cmsFloat64Number) i * 65535.0 / (nResultSamples - 1);
1072
1073 // Find interval in which y is within.
1074 j = GetInterval(y, InCurve->Table16, InCurve->InterpParams);
1075 if (j >= 0) {
1076
1077
1078 // Get limits of interval
1079 x1 = InCurve ->Table16[j];
1080 x2 = InCurve ->Table16[j+1];
1081
1082 y1 = (cmsFloat64Number) (j * 65535.0) / (InCurve ->nEntries - 1);
1083 y2 = (cmsFloat64Number) ((j+1) * 65535.0 ) / (InCurve ->nEntries - 1);
1084
1085 // If collapsed, then use any
1086 if (x1 == x2) {
1087
1088 out ->Table16[i] = _cmsQuickSaturateWord(Ascending ? y2 : y1);
1089 continue;
1090
1091 } else {
1092
1093 // Interpolate
1094 a = (y2 - y1) / (x2 - x1);
1095 b = y2 - a * x2;
1096 }
1097 }
1098
1099 out ->Table16[i] = _cmsQuickSaturateWord(a* y + b);
1100 }
1101
1102
1103 return out;
1104 }
1105
1106 // Reverse a gamma table
1107 cmsToneCurve* CMSEXPORT cmsReverseToneCurve(const cmsToneCurve* InGamma)
1108 {
1109 _cmsAssert(InGamma != NULL);
1110
1111 return cmsReverseToneCurveEx(4096, InGamma);
1112 }
1113
1114 // From: Eilers, P.H.C. (1994) Smoothing and interpolation with finite
1115 // differences. in: Graphic Gems IV, Heckbert, P.S. (ed.), Academic press.
1116 //
1117 // Smoothing and interpolation with second differences.
1118 //
1119 // Input: weights (w), data (y): vector from 1 to m.
1120 // Input: smoothing parameter (lambda), length (m).
1121 // Output: smoothed vector (z): vector from 1 to m.
1122
1123 static
1124 cmsBool smooth2(cmsContext ContextID, cmsFloat32Number w[], cmsFloat32Number y[],
1125 cmsFloat32Number z[], cmsFloat32Number lambda, int m)
1126 {
1127 int i, i1, i2;
1128 cmsFloat32Number *c, *d, *e;
1129 cmsBool st;
1130
1131
1132 c = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1133 d = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1134 e = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1135
1136 if (c != NULL && d != NULL && e != NULL) {
1137
1138
1139 d[1] = w[1] + lambda;
1140 c[1] = -2 * lambda / d[1];
1141 e[1] = lambda /d[1];
1142 z[1] = w[1] * y[1];
1143 d[2] = w[2] + 5 * lambda - d[1] * c[1] * c[1];
1144 c[2] = (-4 * lambda - d[1] * c[1] * e[1]) / d[2];
1145 e[2] = lambda / d[2];
1146 z[2] = w[2] * y[2] - c[1] * z[1];
1147
1148 for (i = 3; i < m - 1; i++) {
1149 i1 = i - 1; i2 = i - 2;
1150 d[i]= w[i] + 6 * lambda - c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1151 c[i] = (-4 * lambda -d[i1] * c[i1] * e[i1])/ d[i];
1152 e[i] = lambda / d[i];
1153 z[i] = w[i] * y[i] - c[i1] * z[i1] - e[i2] * z[i2];
1154 }
1155
1156 i1 = m - 2; i2 = m - 3;
1157
1158 d[m - 1] = w[m - 1] + 5 * lambda -c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1159 c[m - 1] = (-2 * lambda - d[i1] * c[i1] * e[i1]) / d[m - 1];
1160 z[m - 1] = w[m - 1] * y[m - 1] - c[i1] * z[i1] - e[i2] * z[i2];
1161 i1 = m - 1; i2 = m - 2;
1162
1163 d[m] = w[m] + lambda - c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1164 z[m] = (w[m] * y[m] - c[i1] * z[i1] - e[i2] * z[i2]) / d[m];
1165 z[m - 1] = z[m - 1] / d[m - 1] - c[m - 1] * z[m];
1166
1167 for (i = m - 2; 1<= i; i--)
1168 z[i] = z[i] / d[i] - c[i] * z[i + 1] - e[i] * z[i + 2];
1169
1170 st = TRUE;
1171 }
1172 else st = FALSE;
1173
1174 if (c != NULL) _cmsFree(ContextID, c);
1175 if (d != NULL) _cmsFree(ContextID, d);
1176 if (e != NULL) _cmsFree(ContextID, e);
1177
1178 return st;
1179 }
1180
1181 // Smooths a curve sampled at regular intervals.
1182 cmsBool CMSEXPORT cmsSmoothToneCurve(cmsToneCurve* Tab, cmsFloat64Number lambda)
1183 {
1184 cmsBool SuccessStatus = TRUE;
1185 cmsFloat32Number *w, *y, *z;
1186 cmsUInt32Number i, nItems, Zeros, Poles;
1187
1188 if (Tab != NULL && Tab->InterpParams != NULL)
1189 {
1190 cmsContext ContextID = Tab->InterpParams->ContextID;
1191
1192 if (!cmsIsToneCurveLinear(Tab)) // Only non-linear curves need smoothing
1193 {
1194 nItems = Tab->nEntries;
1195 if (nItems < MAX_NODES_IN_CURVE)
1196 {
1197 // Allocate one more item than needed
1198 w = (cmsFloat32Number *)_cmsCalloc(ContextID, nItems + 1, sizeof(cmsFloat32Number));
1199 y = (cmsFloat32Number *)_cmsCalloc(ContextID, nItems + 1, sizeof(cmsFloat32Number));
1200 z = (cmsFloat32Number *)_cmsCalloc(ContextID, nItems + 1, sizeof(cmsFloat32Number));
1201
1202 if (w != NULL && y != NULL && z != NULL) // Ensure no memory allocation failure
1203 {
1204 memset(w, 0, (nItems + 1) * sizeof(cmsFloat32Number));
1205 memset(y, 0, (nItems + 1) * sizeof(cmsFloat32Number));
1206 memset(z, 0, (nItems + 1) * sizeof(cmsFloat32Number));
1207
1208 for (i = 0; i < nItems; i++)
1209 {
1210 y[i + 1] = (cmsFloat32Number)Tab->Table16[i];
1211 w[i + 1] = 1.0;
1212 }
1213
1214 if (smooth2(ContextID, w, y, z, (cmsFloat32Number)lambda, (int)nItems))
1215 {
1216 // Do some reality - checking...
1217
1218 Zeros = Poles = 0;
1219 for (i = nItems; i > 1; --i)
1220 {
1221 if (z[i] == 0.) Zeros++;
1222 if (z[i] >= 65535.) Poles++;
1223 if (z[i] < z[i - 1])
1224 {
1225 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Non-Monotonic.");
1226 SuccessStatus = FALSE;
1227 break;
1228 }
1229 }
1230
1231 if (SuccessStatus && Zeros > (nItems / 3))
1232 {
1233 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Degenerated, mostly zeros.");
1234 SuccessStatus = FALSE;
1235 }
1236
1237 if (SuccessStatus && Poles > (nItems / 3))
1238 {
1239 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Degenerated, mostly poles.");
1240 SuccessStatus = FALSE;
1241 }
1242
1243 if (SuccessStatus) // Seems ok
1244 {
1245 for (i = 0; i < nItems; i++)
1246 {
1247 // Clamp to cmsUInt16Number
1248 Tab->Table16[i] = _cmsQuickSaturateWord(z[i + 1]);
1249 }
1250 }
1251 }
1252 else // Could not smooth
1253 {
1254 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Function smooth2 failed.");
1255 SuccessStatus = FALSE;
1256 }
1257 }
1258 else // One or more buffers could not be allocated
1259 {
1260 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Could not allocate memory.");
1261 SuccessStatus = FALSE;
1262 }
1263
1264 if (z != NULL)
1265 _cmsFree(ContextID, z);
1266
1267 if (y != NULL)
1268 _cmsFree(ContextID, y);
1269
1270 if (w != NULL)
1271 _cmsFree(ContextID, w);
1272 }
1273 else // too many items in the table
1274 {
1275 cmsSignalError(ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Too many points.");
1276 SuccessStatus = FALSE;
1277 }
1278 }
1279 }
1280 else // Tab parameter or Tab->InterpParams is NULL
1281 {
1282 // Can't signal an error here since the ContextID is not known at this point
1283 SuccessStatus = FALSE;
1284 }
1285
1286 return SuccessStatus;
1287 }
1288
1289 // Is a table linear? Do not use parametric since we cannot guarantee some weird parameters resulting
1290 // in a linear table. This way assures it is linear in 12 bits, which should be enought in most cases.
1291 cmsBool CMSEXPORT cmsIsToneCurveLinear(const cmsToneCurve* Curve)
1292 {
1293 int i;
1294 int diff;
1295
1296 _cmsAssert(Curve != NULL);
1297
1298 for (i=0; i < (int) Curve ->nEntries; i++) {
1299
1300 diff = abs((int) Curve->Table16[i] - (int) _cmsQuantizeVal(i, Curve ->nEntries));
1301 if (diff > 0x0f)
1302 return FALSE;
1303 }
1304
1305 return TRUE;
1306 }
1307
1308 // Same, but for monotonicity
1309 cmsBool CMSEXPORT cmsIsToneCurveMonotonic(const cmsToneCurve* t)
1310 {
1311 cmsUInt32Number n;
1312 int i, last;
1313 cmsBool lDescending;
1314
1315 _cmsAssert(t != NULL);
1316
1317 // Degenerated curves are monotonic? Ok, let's pass them
1318 n = t ->nEntries;
1319 if (n < 2) return TRUE;
1320
1321 // Curve direction
1322 lDescending = cmsIsToneCurveDescending(t);
1323
1324 if (lDescending) {
1325
1326 last = t ->Table16[0];
1327
1328 for (i = 1; i < (int) n; i++) {
1329
1330 if (t ->Table16[i] - last > 2) // We allow some ripple
1331 return FALSE;
1332 else
1333 last = t ->Table16[i];
1334
1335 }
1336 }
1337 else {
1338
1339 last = t ->Table16[n-1];
1340
1341 for (i = (int) n - 2; i >= 0; --i) {
1342
1343 if (t ->Table16[i] - last > 2)
1344 return FALSE;
1345 else
1346 last = t ->Table16[i];
1347
1348 }
1349 }
1350
1351 return TRUE;
1352 }
1353
1354 // Same, but for descending tables
1355 cmsBool CMSEXPORT cmsIsToneCurveDescending(const cmsToneCurve* t)
1356 {
1357 _cmsAssert(t != NULL);
1358
1359 return t ->Table16[0] > t ->Table16[t ->nEntries-1];
1360 }
1361
1362
1363 // Another info fn: is out gamma table multisegment?
1364 cmsBool CMSEXPORT cmsIsToneCurveMultisegment(const cmsToneCurve* t)
1365 {
1366 _cmsAssert(t != NULL);
1367
1368 return t -> nSegments > 1;
1369 }
1370
1371 cmsInt32Number CMSEXPORT cmsGetToneCurveParametricType(const cmsToneCurve* t)
1372 {
1373 _cmsAssert(t != NULL);
1374
1375 if (t -> nSegments != 1) return 0;
1376 return t ->Segments[0].Type;
1377 }
1378
1379 // We need accuracy this time
1380 cmsFloat32Number CMSEXPORT cmsEvalToneCurveFloat(const cmsToneCurve* Curve, cmsFloat32Number v)
1381 {
1382 _cmsAssert(Curve != NULL);
1383
1384 // Check for 16 bits table. If so, this is a limited-precision tone curve
1385 if (Curve ->nSegments == 0) {
1386
1387 cmsUInt16Number In, Out;
1388
1389 In = (cmsUInt16Number) _cmsQuickSaturateWord(v * 65535.0);
1390 Out = cmsEvalToneCurve16(Curve, In);
1391
1392 return (cmsFloat32Number) (Out / 65535.0);
1393 }
1394
1395 return (cmsFloat32Number) EvalSegmentedFn(Curve, v);
1396 }
1397
1398 // We need xput over here
1399 cmsUInt16Number CMSEXPORT cmsEvalToneCurve16(const cmsToneCurve* Curve, cmsUInt16Number v)
1400 {
1401 cmsUInt16Number out;
1402
1403 _cmsAssert(Curve != NULL);
1404
1405 Curve ->InterpParams ->Interpolation.Lerp16(&v, &out, Curve ->InterpParams);
1406 return out;
1407 }
1408
1409
1410 // Least squares fitting.
1411 // A mathematical procedure for finding the best-fitting curve to a given set of points by
1412 // minimizing the sum of the squares of the offsets ("the residuals") of the points from the curve.
1413 // The sum of the squares of the offsets is used instead of the offset absolute values because
1414 // this allows the residuals to be treated as a continuous differentiable quantity.
1415 //
1416 // y = f(x) = x ^ g
1417 //
1418 // R = (yi - (xi^g))
1419 // R2 = (yi - (xi^g))2
1420 // SUM R2 = SUM (yi - (xi^g))2
1421 //
1422 // dR2/dg = -2 SUM x^g log(x)(y - x^g)
1423 // solving for dR2/dg = 0
1424 //
1425 // g = 1/n * SUM(log(y) / log(x))
1426
1427 cmsFloat64Number CMSEXPORT cmsEstimateGamma(const cmsToneCurve* t, cmsFloat64Number Precision)
1428 {
1429 cmsFloat64Number gamma, sum, sum2;
1430 cmsFloat64Number n, x, y, Std;
1431 cmsUInt32Number i;
1432
1433 _cmsAssert(t != NULL);
1434
1435 sum = sum2 = n = 0;
1436
1437 // Excluding endpoints
1438 for (i=1; i < (MAX_NODES_IN_CURVE-1); i++) {
1439
1440 x = (cmsFloat64Number) i / (MAX_NODES_IN_CURVE-1);
1441 y = (cmsFloat64Number) cmsEvalToneCurveFloat(t, (cmsFloat32Number) x);
1442
1443 // Avoid 7% on lower part to prevent
1444 // artifacts due to linear ramps
1445
1446 if (y > 0. && y < 1. && x > 0.07) {
1447
1448 gamma = log(y) / log(x);
1449 sum += gamma;
1450 sum2 += gamma * gamma;
1451 n++;
1452 }
1453 }
1454
1455 // Take a look on SD to see if gamma isn't exponential at all
1456 Std = sqrt((n * sum2 - sum * sum) / (n*(n-1)));
1457
1458 if (Std > Precision)
1459 return -1.0;
1460
1461 return (sum / n); // The mean
1462 }