1 //---------------------------------------------------------------------------------
2 //
3 // Little Color Management System
4 // Copyright (c) 1998-2013 Marti Maria Saguer
5 //
6 // Permission is hereby granted, free of charge, to any person obtaining
7 // a copy of this software and associated documentation files (the "Software"),
8 // to deal in the Software without restriction, including without limitation
9 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 // and/or sell copies of the Software, and to permit persons to whom the Software
11 // is furnished to do so, subject to the following conditions:
12 //
13 // The above copyright notice and this permission notice shall be included in
14 // all copies or substantial portions of the Software.
15 //
16 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
17 // EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
18 // THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
19 // NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
20 // LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
21 // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
22 // WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
23 //
24 //---------------------------------------------------------------------------------
25 //
26
27 #include "lcms2_internal.h"
28
29 // Tone curves are powerful constructs that can contain curves specified in diverse ways.
30 // The curve is stored in segments, where each segment can be sampled or specified by parameters.
31 // a 16.bit simplification of the *whole* curve is kept for optimization purposes. For float operation,
32 // each segment is evaluated separately. Plug-ins may be used to define new parametric schemes,
33 // each plug-in may define up to MAX_TYPES_IN_LCMS_PLUGIN functions types. For defining a function,
34 // the plug-in should provide the type id, how many parameters each type has, and a pointer to
35 // a procedure that evaluates the function. In the case of reverse evaluation, the evaluator will
36 // be called with the type id as a negative value, and a sampled version of the reversed curve
37 // will be built.
38
39 // ----------------------------------------------------------------- Implementation
40 // Maxim number of nodes
41 #define MAX_NODES_IN_CURVE 4097
42 #define MINUS_INF (-1E22F)
43 #define PLUS_INF (+1E22F)
44
45 // The list of supported parametric curves
46 typedef struct _cmsParametricCurvesCollection_st {
47
48 int nFunctions; // Number of supported functions in this chunk
49 int FunctionTypes[MAX_TYPES_IN_LCMS_PLUGIN]; // The identification types
50 int ParameterCount[MAX_TYPES_IN_LCMS_PLUGIN]; // Number of parameters for each function
51 cmsParametricCurveEvaluator Evaluator; // The evaluator
52
53 struct _cmsParametricCurvesCollection_st* Next; // Next in list
54
55 } _cmsParametricCurvesCollection;
56
57 // This is the default (built-in) evaluator
58 static cmsFloat64Number DefaultEvalParametricFn(cmsInt32Number Type, const cmsFloat64Number Params[], cmsFloat64Number R);
59
60 // The built-in list
61 static _cmsParametricCurvesCollection DefaultCurves = {
62 9, // # of curve types
63 { 1, 2, 3, 4, 5, 6, 7, 8, 108 }, // Parametric curve ID
64 { 1, 3, 4, 5, 7, 4, 5, 5, 1 }, // Parameters by type
65 DefaultEvalParametricFn, // Evaluator
66 NULL // Next in chain
67 };
68
69 // Duplicates the zone of memory used by the plug-in in the new context
70 static
DupPluginCurvesList(struct _cmsContext_struct * ctx,const struct _cmsContext_struct * src)71 void DupPluginCurvesList(struct _cmsContext_struct* ctx,
72 const struct _cmsContext_struct* src)
73 {
74 _cmsCurvesPluginChunkType newHead = { NULL };
75 _cmsParametricCurvesCollection* entry;
76 _cmsParametricCurvesCollection* Anterior = NULL;
77 _cmsCurvesPluginChunkType* head = (_cmsCurvesPluginChunkType*) src->chunks[CurvesPlugin];
78
79 _cmsAssert(head != NULL);
80
81 // Walk the list copying all nodes
82 for (entry = head->ParametricCurves;
83 entry != NULL;
84 entry = entry ->Next) {
85
86 _cmsParametricCurvesCollection *newEntry = ( _cmsParametricCurvesCollection *) _cmsSubAllocDup(ctx ->MemPool, entry, sizeof(_cmsParametricCurvesCollection));
87
88 if (newEntry == NULL)
89 return;
90
91 // We want to keep the linked list order, so this is a little bit tricky
92 newEntry -> Next = NULL;
93 if (Anterior)
94 Anterior -> Next = newEntry;
95
96 Anterior = newEntry;
97
98 if (newHead.ParametricCurves == NULL)
99 newHead.ParametricCurves = newEntry;
100 }
101
102 ctx ->chunks[CurvesPlugin] = _cmsSubAllocDup(ctx->MemPool, &newHead, sizeof(_cmsCurvesPluginChunkType));
103 }
104
105 // The allocator have to follow the chain
_cmsAllocCurvesPluginChunk(struct _cmsContext_struct * ctx,const struct _cmsContext_struct * src)106 void _cmsAllocCurvesPluginChunk(struct _cmsContext_struct* ctx,
107 const struct _cmsContext_struct* src)
108 {
109 _cmsAssert(ctx != NULL);
110
111 if (src != NULL) {
112
113 // Copy all linked list
114 DupPluginCurvesList(ctx, src);
115 }
116 else {
117 static _cmsCurvesPluginChunkType CurvesPluginChunk = { NULL };
118 ctx ->chunks[CurvesPlugin] = _cmsSubAllocDup(ctx ->MemPool, &CurvesPluginChunk, sizeof(_cmsCurvesPluginChunkType));
119 }
120 }
121
122
123 // The linked list head
124 _cmsCurvesPluginChunkType _cmsCurvesPluginChunk = { NULL };
125
126 // As a way to install new parametric curves
_cmsRegisterParametricCurvesPlugin(cmsContext ContextID,cmsPluginBase * Data)127 cmsBool _cmsRegisterParametricCurvesPlugin(cmsContext ContextID, cmsPluginBase* Data)
128 {
129 _cmsCurvesPluginChunkType* ctx = ( _cmsCurvesPluginChunkType*) _cmsContextGetClientChunk(ContextID, CurvesPlugin);
130 cmsPluginParametricCurves* Plugin = (cmsPluginParametricCurves*) Data;
131 _cmsParametricCurvesCollection* fl;
132
133 if (Data == NULL) {
134
135 ctx -> ParametricCurves = NULL;
136 return TRUE;
137 }
138
139 fl = (_cmsParametricCurvesCollection*) _cmsPluginMalloc(ContextID, sizeof(_cmsParametricCurvesCollection));
140 if (fl == NULL) return FALSE;
141
142 // Copy the parameters
143 fl ->Evaluator = Plugin ->Evaluator;
144 fl ->nFunctions = Plugin ->nFunctions;
145
146 // Make sure no mem overwrites
147 if (fl ->nFunctions > MAX_TYPES_IN_LCMS_PLUGIN)
148 fl ->nFunctions = MAX_TYPES_IN_LCMS_PLUGIN;
149
150 // Copy the data
151 memmove(fl->FunctionTypes, Plugin ->FunctionTypes, fl->nFunctions * sizeof(cmsUInt32Number));
152 memmove(fl->ParameterCount, Plugin ->ParameterCount, fl->nFunctions * sizeof(cmsUInt32Number));
153
154 // Keep linked list
155 fl ->Next = ctx->ParametricCurves;
156 ctx->ParametricCurves = fl;
157
158 // All is ok
159 return TRUE;
160 }
161
162
163 // Search in type list, return position or -1 if not found
164 static
IsInSet(int Type,_cmsParametricCurvesCollection * c)165 int IsInSet(int Type, _cmsParametricCurvesCollection* c)
166 {
167 int i;
168
169 for (i=0; i < c ->nFunctions; i++)
170 if (abs(Type) == c ->FunctionTypes[i]) return i;
171
172 return -1;
173 }
174
175
176 // Search for the collection which contains a specific type
177 static
GetParametricCurveByType(cmsContext ContextID,int Type,int * index)178 _cmsParametricCurvesCollection *GetParametricCurveByType(cmsContext ContextID, int Type, int* index)
179 {
180 _cmsParametricCurvesCollection* c;
181 int Position;
182 _cmsCurvesPluginChunkType* ctx = ( _cmsCurvesPluginChunkType*) _cmsContextGetClientChunk(ContextID, CurvesPlugin);
183
184 for (c = ctx->ParametricCurves; c != NULL; c = c ->Next) {
185
186 Position = IsInSet(Type, c);
187
188 if (Position != -1) {
189 if (index != NULL)
190 *index = Position;
191 return c;
192 }
193 }
194 // If none found, revert for defaults
195 for (c = &DefaultCurves; c != NULL; c = c ->Next) {
196
197 Position = IsInSet(Type, c);
198
199 if (Position != -1) {
200 if (index != NULL)
201 *index = Position;
202 return c;
203 }
204 }
205
206 return NULL;
207 }
208
209 // Low level allocate, which takes care of memory details. nEntries may be zero, and in this case
210 // no optimation curve is computed. nSegments may also be zero in the inverse case, where only the
211 // optimization curve is given. Both features simultaneously is an error
212 static
AllocateToneCurveStruct(cmsContext ContextID,cmsInt32Number nEntries,cmsInt32Number nSegments,const cmsCurveSegment * Segments,const cmsUInt16Number * Values)213 cmsToneCurve* AllocateToneCurveStruct(cmsContext ContextID, cmsInt32Number nEntries,
214 cmsInt32Number nSegments, const cmsCurveSegment* Segments,
215 const cmsUInt16Number* Values)
216 {
217 cmsToneCurve* p;
218 int i;
219
220 // We allow huge tables, which are then restricted for smoothing operations
221 if (nEntries > 65530 || nEntries < 0) {
222 cmsSignalError(ContextID, cmsERROR_RANGE, "Couldn't create tone curve of more than 65530 entries");
223 return NULL;
224 }
225
226 if (nEntries <= 0 && nSegments <= 0) {
227 cmsSignalError(ContextID, cmsERROR_RANGE, "Couldn't create tone curve with zero segments and no table");
228 return NULL;
229 }
230
231 // Allocate all required pointers, etc.
232 p = (cmsToneCurve*) _cmsMallocZero(ContextID, sizeof(cmsToneCurve));
233 if (!p) return NULL;
234
235 // In this case, there are no segments
236 if (nSegments <= 0) {
237 p ->Segments = NULL;
238 p ->Evals = NULL;
239 }
240 else {
241 p ->Segments = (cmsCurveSegment*) _cmsCalloc(ContextID, nSegments, sizeof(cmsCurveSegment));
242 if (p ->Segments == NULL) goto Error;
243
244 p ->Evals = (cmsParametricCurveEvaluator*) _cmsCalloc(ContextID, nSegments, sizeof(cmsParametricCurveEvaluator));
245 if (p ->Evals == NULL) goto Error;
246 }
247
248 p -> nSegments = nSegments;
249
250 // This 16-bit table contains a limited precision representation of the whole curve and is kept for
251 // increasing xput on certain operations.
252 if (nEntries <= 0) {
253 p ->Table16 = NULL;
254 }
255 else {
256 p ->Table16 = (cmsUInt16Number*) _cmsCalloc(ContextID, nEntries, sizeof(cmsUInt16Number));
257 if (p ->Table16 == NULL) goto Error;
258 }
259
260 p -> nEntries = nEntries;
261
262 // Initialize members if requested
263 if (Values != NULL && (nEntries > 0)) {
264
265 for (i=0; i < nEntries; i++)
266 p ->Table16[i] = Values[i];
267 }
268
269 // Initialize the segments stuff. The evaluator for each segment is located and a pointer to it
270 // is placed in advance to maximize performance.
271 if (Segments != NULL && (nSegments > 0)) {
272
273 _cmsParametricCurvesCollection *c;
274
275 p ->SegInterp = (cmsInterpParams**) _cmsCalloc(ContextID, nSegments, sizeof(cmsInterpParams*));
276 if (p ->SegInterp == NULL) goto Error;
277
278 for (i=0; i< nSegments; i++) {
279
280 // Type 0 is a special marker for table-based curves
281 if (Segments[i].Type == 0)
282 p ->SegInterp[i] = _cmsComputeInterpParams(ContextID, Segments[i].nGridPoints, 1, 1, NULL, CMS_LERP_FLAGS_FLOAT);
283
284 memmove(&p ->Segments[i], &Segments[i], sizeof(cmsCurveSegment));
285
286 if (Segments[i].Type == 0 && Segments[i].SampledPoints != NULL)
287 p ->Segments[i].SampledPoints = (cmsFloat32Number*) _cmsDupMem(ContextID, Segments[i].SampledPoints, sizeof(cmsFloat32Number) * Segments[i].nGridPoints);
288 else
289 p ->Segments[i].SampledPoints = NULL;
290
291
292 c = GetParametricCurveByType(ContextID, Segments[i].Type, NULL);
293 if (c != NULL)
294 p ->Evals[i] = c ->Evaluator;
295 }
296 }
297
298 p ->InterpParams = _cmsComputeInterpParams(ContextID, p ->nEntries, 1, 1, p->Table16, CMS_LERP_FLAGS_16BITS);
299 if (p->InterpParams != NULL)
300 return p;
301
302 Error:
303 if (p -> Segments) _cmsFree(ContextID, p ->Segments);
304 if (p -> Evals) _cmsFree(ContextID, p -> Evals);
305 if (p ->Table16) _cmsFree(ContextID, p ->Table16);
306 _cmsFree(ContextID, p);
307 return NULL;
308 }
309
310
311 // Parametric Fn using floating point
312 static
DefaultEvalParametricFn(cmsInt32Number Type,const cmsFloat64Number Params[],cmsFloat64Number R)313 cmsFloat64Number DefaultEvalParametricFn(cmsInt32Number Type, const cmsFloat64Number Params[], cmsFloat64Number R)
314 {
315 cmsFloat64Number e, Val, disc;
316
317 switch (Type) {
318
319 // X = Y ^ Gamma
320 case 1:
321 if (R < 0) {
322
323 if (fabs(Params[0] - 1.0) < MATRIX_DET_TOLERANCE)
324 Val = R;
325 else
326 Val = 0;
327 }
328 else
329 Val = pow(R, Params[0]);
330 break;
331
332 // Type 1 Reversed: X = Y ^1/gamma
333 case -1:
334 if (R < 0) {
335
336 if (fabs(Params[0] - 1.0) < MATRIX_DET_TOLERANCE)
337 Val = R;
338 else
339 Val = 0;
340 }
341 else
342 Val = pow(R, 1/Params[0]);
343 break;
344
345 // CIE 122-1966
346 // Y = (aX + b)^Gamma | X >= -b/a
347 // Y = 0 | else
348 case 2:
349 disc = -Params[2] / Params[1];
350
351 if (R >= disc ) {
352
353 e = Params[1]*R + Params[2];
354
355 if (e > 0)
356 Val = pow(e, Params[0]);
357 else
358 Val = 0;
359 }
360 else
361 Val = 0;
362 break;
363
364 // Type 2 Reversed
365 // X = (Y ^1/g - b) / a
366 case -2:
367 if (R < 0)
368 Val = 0;
369 else
370 Val = (pow(R, 1.0/Params[0]) - Params[2]) / Params[1];
371
372 if (Val < 0)
373 Val = 0;
374 break;
375
376
377 // IEC 61966-3
378 // Y = (aX + b)^Gamma | X <= -b/a
379 // Y = c | else
380 case 3:
381 disc = -Params[2] / Params[1];
382 if (disc < 0)
383 disc = 0;
384
385 if (R >= disc) {
386
387 e = Params[1]*R + Params[2];
388
389 if (e > 0)
390 Val = pow(e, Params[0]) + Params[3];
391 else
392 Val = 0;
393 }
394 else
395 Val = Params[3];
396 break;
397
398
399 // Type 3 reversed
400 // X=((Y-c)^1/g - b)/a | (Y>=c)
401 // X=-b/a | (Y<c)
402 case -3:
403 if (R >= Params[3]) {
404
405 e = R - Params[3];
406
407 if (e > 0)
408 Val = (pow(e, 1/Params[0]) - Params[2]) / Params[1];
409 else
410 Val = 0;
411 }
412 else {
413 Val = -Params[2] / Params[1];
414 }
415 break;
416
417
418 // IEC 61966-2.1 (sRGB)
419 // Y = (aX + b)^Gamma | X >= d
420 // Y = cX | X < d
421 case 4:
422 if (R >= Params[4]) {
423
424 e = Params[1]*R + Params[2];
425
426 if (e > 0)
427 Val = pow(e, Params[0]);
428 else
429 Val = 0;
430 }
431 else
432 Val = R * Params[3];
433 break;
434
435 // Type 4 reversed
436 // X=((Y^1/g-b)/a) | Y >= (ad+b)^g
437 // X=Y/c | Y< (ad+b)^g
438 case -4:
439 e = Params[1] * Params[4] + Params[2];
440 if (e < 0)
441 disc = 0;
442 else
443 disc = pow(e, Params[0]);
444
445 if (R >= disc) {
446
447 Val = (pow(R, 1.0/Params[0]) - Params[2]) / Params[1];
448 }
449 else {
450 Val = R / Params[3];
451 }
452 break;
453
454
455 // Y = (aX + b)^Gamma + e | X >= d
456 // Y = cX + f | X < d
457 case 5:
458 if (R >= Params[4]) {
459
460 e = Params[1]*R + Params[2];
461
462 if (e > 0)
463 Val = pow(e, Params[0]) + Params[5];
464 else
465 Val = Params[5];
466 }
467 else
468 Val = R*Params[3] + Params[6];
469 break;
470
471
472 // Reversed type 5
473 // X=((Y-e)1/g-b)/a | Y >=(ad+b)^g+e), cd+f
474 // X=(Y-f)/c | else
475 case -5:
476
477 disc = Params[3] * Params[4] + Params[6];
478 if (R >= disc) {
479
480 e = R - Params[5];
481 if (e < 0)
482 Val = 0;
483 else
484 Val = (pow(e, 1.0/Params[0]) - Params[2]) / Params[1];
485 }
486 else {
487 Val = (R - Params[6]) / Params[3];
488 }
489 break;
490
491
492 // Types 6,7,8 comes from segmented curves as described in ICCSpecRevision_02_11_06_Float.pdf
493 // Type 6 is basically identical to type 5 without d
494
495 // Y = (a * X + b) ^ Gamma + c
496 case 6:
497 e = Params[1]*R + Params[2];
498
499 if (e < 0)
500 Val = Params[3];
501 else
502 Val = pow(e, Params[0]) + Params[3];
503 break;
504
505 // ((Y - c) ^1/Gamma - b) / a
506 case -6:
507 e = R - Params[3];
508 if (e < 0)
509 Val = 0;
510 else
511 Val = (pow(e, 1.0/Params[0]) - Params[2]) / Params[1];
512 break;
513
514
515 // Y = a * log (b * X^Gamma + c) + d
516 case 7:
517
518 e = Params[2] * pow(R, Params[0]) + Params[3];
519 if (e <= 0)
520 Val = Params[4];
521 else
522 Val = Params[1]*log10(e) + Params[4];
523 break;
524
525 // (Y - d) / a = log(b * X ^Gamma + c)
526 // pow(10, (Y-d) / a) = b * X ^Gamma + c
527 // pow((pow(10, (Y-d) / a) - c) / b, 1/g) = X
528 case -7:
529 Val = pow((pow(10.0, (R-Params[4]) / Params[1]) - Params[3]) / Params[2], 1.0 / Params[0]);
530 break;
531
532
533 //Y = a * b^(c*X+d) + e
534 case 8:
535 Val = (Params[0] * pow(Params[1], Params[2] * R + Params[3]) + Params[4]);
536 break;
537
538
539 // Y = (log((y-e) / a) / log(b) - d ) / c
540 // a=0, b=1, c=2, d=3, e=4,
541 case -8:
542
543 disc = R - Params[4];
544 if (disc < 0) Val = 0;
545 else
546 Val = (log(disc / Params[0]) / log(Params[1]) - Params[3]) / Params[2];
547 break;
548
549 // S-Shaped: (1 - (1-x)^1/g)^1/g
550 case 108:
551 Val = pow(1.0 - pow(1 - R, 1/Params[0]), 1/Params[0]);
552 break;
553
554 // y = (1 - (1-x)^1/g)^1/g
555 // y^g = (1 - (1-x)^1/g)
556 // 1 - y^g = (1-x)^1/g
557 // (1 - y^g)^g = 1 - x
558 // 1 - (1 - y^g)^g
559 case -108:
560 Val = 1 - pow(1 - pow(R, Params[0]), Params[0]);
561 break;
562
563 default:
564 // Unsupported parametric curve. Should never reach here
565 return 0;
566 }
567
568 return Val;
569 }
570
571 // Evaluate a segmented funtion for a single value. Return -1 if no valid segment found .
572 // If fn type is 0, perform an interpolation on the table
573 static
EvalSegmentedFn(const cmsToneCurve * g,cmsFloat64Number R)574 cmsFloat64Number EvalSegmentedFn(const cmsToneCurve *g, cmsFloat64Number R)
575 {
576 int i;
577
578 for (i = g ->nSegments-1; i >= 0 ; --i) {
579
580 // Check for domain
581 if ((R > g ->Segments[i].x0) && (R <= g ->Segments[i].x1)) {
582
583 // Type == 0 means segment is sampled
584 if (g ->Segments[i].Type == 0) {
585
586 cmsFloat32Number R1 = (cmsFloat32Number) (R - g ->Segments[i].x0) / (g ->Segments[i].x1 - g ->Segments[i].x0);
587 cmsFloat32Number Out;
588
589 // Setup the table (TODO: clean that)
590 g ->SegInterp[i]-> Table = g ->Segments[i].SampledPoints;
591
592 g ->SegInterp[i] -> Interpolation.LerpFloat(&R1, &Out, g ->SegInterp[i]);
593
594 return Out;
595 }
596 else
597 return g ->Evals[i](g->Segments[i].Type, g ->Segments[i].Params, R);
598 }
599 }
600
601 return MINUS_INF;
602 }
603
604 // Access to estimated low-res table
cmsGetToneCurveEstimatedTableEntries(const cmsToneCurve * t)605 cmsUInt32Number CMSEXPORT cmsGetToneCurveEstimatedTableEntries(const cmsToneCurve* t)
606 {
607 _cmsAssert(t != NULL);
608 return t ->nEntries;
609 }
610
cmsGetToneCurveEstimatedTable(const cmsToneCurve * t)611 const cmsUInt16Number* CMSEXPORT cmsGetToneCurveEstimatedTable(const cmsToneCurve* t)
612 {
613 _cmsAssert(t != NULL);
614 return t ->Table16;
615 }
616
617
618 // Create an empty gamma curve, by using tables. This specifies only the limited-precision part, and leaves the
619 // floating point description empty.
cmsBuildTabulatedToneCurve16(cmsContext ContextID,cmsInt32Number nEntries,const cmsUInt16Number Values[])620 cmsToneCurve* CMSEXPORT cmsBuildTabulatedToneCurve16(cmsContext ContextID, cmsInt32Number nEntries, const cmsUInt16Number Values[])
621 {
622 return AllocateToneCurveStruct(ContextID, nEntries, 0, NULL, Values);
623 }
624
625 static
EntriesByGamma(cmsFloat64Number Gamma)626 int EntriesByGamma(cmsFloat64Number Gamma)
627 {
628 if (fabs(Gamma - 1.0) < 0.001) return 2;
629 return 4096;
630 }
631
632
633 // Create a segmented gamma, fill the table
cmsBuildSegmentedToneCurve(cmsContext ContextID,cmsInt32Number nSegments,const cmsCurveSegment Segments[])634 cmsToneCurve* CMSEXPORT cmsBuildSegmentedToneCurve(cmsContext ContextID,
635 cmsInt32Number nSegments, const cmsCurveSegment Segments[])
636 {
637 int i;
638 cmsFloat64Number R, Val;
639 cmsToneCurve* g;
640 int nGridPoints = 4096;
641
642 _cmsAssert(Segments != NULL);
643
644 // Optimizatin for identity curves.
645 if (nSegments == 1 && Segments[0].Type == 1) {
646
647 nGridPoints = EntriesByGamma(Segments[0].Params[0]);
648 }
649
650 g = AllocateToneCurveStruct(ContextID, nGridPoints, nSegments, Segments, NULL);
651 if (g == NULL) return NULL;
652
653 // Once we have the floating point version, we can approximate a 16 bit table of 4096 entries
654 // for performance reasons. This table would normally not be used except on 8/16 bits transforms.
655 for (i=0; i < nGridPoints; i++) {
656
657 R = (cmsFloat64Number) i / (nGridPoints-1);
658
659 Val = EvalSegmentedFn(g, R);
660
661 // Round and saturate
662 g ->Table16[i] = _cmsQuickSaturateWord(Val * 65535.0);
663 }
664
665 return g;
666 }
667
668 // Use a segmented curve to store the floating point table
cmsBuildTabulatedToneCurveFloat(cmsContext ContextID,cmsUInt32Number nEntries,const cmsFloat32Number values[])669 cmsToneCurve* CMSEXPORT cmsBuildTabulatedToneCurveFloat(cmsContext ContextID, cmsUInt32Number nEntries, const cmsFloat32Number values[])
670 {
671 cmsCurveSegment Seg[3];
672
673 // A segmented tone curve should have function segments in the first and last positions
674 // Initialize segmented curve part up to 0 to constant value = samples[0]
675 Seg[0].x0 = MINUS_INF;
676 Seg[0].x1 = 0;
677 Seg[0].Type = 6;
678
679 Seg[0].Params[0] = 1;
680 Seg[0].Params[1] = 0;
681 Seg[0].Params[2] = 0;
682 Seg[0].Params[3] = values[0];
683 Seg[0].Params[4] = 0;
684
685 // From zero to 1
686 Seg[1].x0 = 0;
687 Seg[1].x1 = 1.0;
688 Seg[1].Type = 0;
689
690 Seg[1].nGridPoints = nEntries;
691 Seg[1].SampledPoints = (cmsFloat32Number*) values;
692
693 // Final segment is constant = lastsample
694 Seg[2].x0 = 1.0;
695 Seg[2].x1 = PLUS_INF;
696 Seg[2].Type = 6;
697
698 Seg[2].Params[0] = 1;
699 Seg[2].Params[1] = 0;
700 Seg[2].Params[2] = 0;
701 Seg[2].Params[3] = values[nEntries-1];
702 Seg[2].Params[4] = 0;
703
704
705 return cmsBuildSegmentedToneCurve(ContextID, 3, Seg);
706 }
707
708 // Parametric curves
709 //
710 // Parameters goes as: Curve, a, b, c, d, e, f
711 // Type is the ICC type +1
712 // if type is negative, then the curve is analyticaly inverted
cmsBuildParametricToneCurve(cmsContext ContextID,cmsInt32Number Type,const cmsFloat64Number Params[])713 cmsToneCurve* CMSEXPORT cmsBuildParametricToneCurve(cmsContext ContextID, cmsInt32Number Type, const cmsFloat64Number Params[])
714 {
715 cmsCurveSegment Seg0;
716 int Pos = 0;
717 cmsUInt32Number size;
718 _cmsParametricCurvesCollection* c = GetParametricCurveByType(ContextID, Type, &Pos);
719
720 _cmsAssert(Params != NULL);
721
722 if (c == NULL) {
723 cmsSignalError(ContextID, cmsERROR_UNKNOWN_EXTENSION, "Invalid parametric curve type %d", Type);
724 return NULL;
725 }
726
727 memset(&Seg0, 0, sizeof(Seg0));
728
729 Seg0.x0 = MINUS_INF;
730 Seg0.x1 = PLUS_INF;
731 Seg0.Type = Type;
732
733 size = c->ParameterCount[Pos] * sizeof(cmsFloat64Number);
734 memmove(Seg0.Params, Params, size);
735
736 return cmsBuildSegmentedToneCurve(ContextID, 1, &Seg0);
737 }
738
739
740
741 // Build a gamma table based on gamma constant
cmsBuildGamma(cmsContext ContextID,cmsFloat64Number Gamma)742 cmsToneCurve* CMSEXPORT cmsBuildGamma(cmsContext ContextID, cmsFloat64Number Gamma)
743 {
744 return cmsBuildParametricToneCurve(ContextID, 1, &Gamma);
745 }
746
747
748 // Free all memory taken by the gamma curve
cmsFreeToneCurve(cmsToneCurve * Curve)749 void CMSEXPORT cmsFreeToneCurve(cmsToneCurve* Curve)
750 {
751 cmsContext ContextID;
752
753 // added by Xiaochuan Liu
754 // Curve->InterpParams may be null
755 if (Curve == NULL || Curve->InterpParams == NULL) return;
756
757 ContextID = Curve ->InterpParams->ContextID;
758
759 _cmsFreeInterpParams(Curve ->InterpParams);
760 Curve ->InterpParams = NULL;
761
762 if (Curve -> Table16)
763 {
764 _cmsFree(ContextID, Curve ->Table16);
765 Curve ->Table16 = NULL;
766 }
767
768 if (Curve ->Segments) {
769
770 cmsUInt32Number i;
771
772 for (i=0; i < Curve ->nSegments; i++) {
773
774 if (Curve ->Segments[i].SampledPoints) {
775 _cmsFree(ContextID, Curve ->Segments[i].SampledPoints);
776 Curve ->Segments[i].SampledPoints = NULL;
777 }
778
779 if (Curve ->SegInterp[i] != 0)
780 {
781 _cmsFreeInterpParams(Curve->SegInterp[i]);
782 Curve->SegInterp[i] = NULL;
783 }
784 }
785
786 _cmsFree(ContextID, Curve ->Segments);
787 Curve ->Segments = NULL;
788 _cmsFree(ContextID, Curve ->SegInterp);
789 Curve ->SegInterp = NULL;
790 }
791
792 if (Curve -> Evals)
793 {
794 _cmsFree(ContextID, Curve -> Evals);
795 Curve -> Evals = NULL;
796 }
797
798 if (Curve)
799 {
800 _cmsFree(ContextID, Curve);
801 Curve = NULL;
802 }
803 }
804
805 // Utility function, free 3 gamma tables
cmsFreeToneCurveTriple(cmsToneCurve * Curve[3])806 void CMSEXPORT cmsFreeToneCurveTriple(cmsToneCurve* Curve[3])
807 {
808
809 _cmsAssert(Curve != NULL);
810
811 if (Curve[0] != NULL) cmsFreeToneCurve(Curve[0]);
812 if (Curve[1] != NULL) cmsFreeToneCurve(Curve[1]);
813 if (Curve[2] != NULL) cmsFreeToneCurve(Curve[2]);
814
815 Curve[0] = Curve[1] = Curve[2] = NULL;
816 }
817
818
819 // Duplicate a gamma table
cmsDupToneCurve(const cmsToneCurve * In)820 cmsToneCurve* CMSEXPORT cmsDupToneCurve(const cmsToneCurve* In)
821 {
822 // Xiaochuan Liu
823 // fix openpdf bug(mantis id:0055683, google id:360198)
824 // the function CurveSetElemTypeFree in cmslut.c also needs to check pointer
825 if (In == NULL || In ->InterpParams == NULL) return NULL;
826
827 return AllocateToneCurveStruct(In ->InterpParams ->ContextID, In ->nEntries, In ->nSegments, In ->Segments, In ->Table16);
828 }
829
830 // Joins two curves for X and Y. Curves should be monotonic.
831 // We want to get
832 //
833 // y = Y^-1(X(t))
834 //
cmsJoinToneCurve(cmsContext ContextID,const cmsToneCurve * X,const cmsToneCurve * Y,cmsUInt32Number nResultingPoints)835 cmsToneCurve* CMSEXPORT cmsJoinToneCurve(cmsContext ContextID,
836 const cmsToneCurve* X,
837 const cmsToneCurve* Y, cmsUInt32Number nResultingPoints)
838 {
839 cmsToneCurve* out = NULL;
840 cmsToneCurve* Yreversed = NULL;
841 cmsFloat32Number t, x;
842 cmsFloat32Number* Res = NULL;
843 cmsUInt32Number i;
844
845
846 _cmsAssert(X != NULL);
847 _cmsAssert(Y != NULL);
848
849 Yreversed = cmsReverseToneCurveEx(nResultingPoints, Y);
850 if (Yreversed == NULL) goto Error;
851
852 Res = (cmsFloat32Number*) _cmsCalloc(ContextID, nResultingPoints, sizeof(cmsFloat32Number));
853 if (Res == NULL) goto Error;
854
855 //Iterate
856 for (i=0; i < nResultingPoints; i++) {
857
858 t = (cmsFloat32Number) i / (nResultingPoints-1);
859 x = cmsEvalToneCurveFloat(X, t);
860 Res[i] = cmsEvalToneCurveFloat(Yreversed, x);
861 }
862
863 // Allocate space for output
864 out = cmsBuildTabulatedToneCurveFloat(ContextID, nResultingPoints, Res);
865
866 Error:
867
868 if (Res != NULL) _cmsFree(ContextID, Res);
869 if (Yreversed != NULL) cmsFreeToneCurve(Yreversed);
870
871 return out;
872 }
873
874
875
876 // Get the surrounding nodes. This is tricky on non-monotonic tables
877 static
GetInterval(cmsFloat64Number In,const cmsUInt16Number LutTable[],const struct _cms_interp_struc * p)878 int GetInterval(cmsFloat64Number In, const cmsUInt16Number LutTable[], const struct _cms_interp_struc* p)
879 {
880 int i;
881 int y0, y1;
882
883 // A 1 point table is not allowed
884 if (p -> Domain[0] < 1) return -1;
885
886 // Let's see if ascending or descending.
887 if (LutTable[0] < LutTable[p ->Domain[0]]) {
888
889 // Table is overall ascending
890 for (i=p->Domain[0]-1; i >=0; --i) {
891
892 y0 = LutTable[i];
893 y1 = LutTable[i+1];
894
895 if (y0 <= y1) { // Increasing
896 if (In >= y0 && In <= y1) return i;
897 }
898 else
899 if (y1 < y0) { // Decreasing
900 if (In >= y1 && In <= y0) return i;
901 }
902 }
903 }
904 else {
905 // Table is overall descending
906 for (i=0; i < (int) p -> Domain[0]; i++) {
907
908 y0 = LutTable[i];
909 y1 = LutTable[i+1];
910
911 if (y0 <= y1) { // Increasing
912 if (In >= y0 && In <= y1) return i;
913 }
914 else
915 if (y1 < y0) { // Decreasing
916 if (In >= y1 && In <= y0) return i;
917 }
918 }
919 }
920
921 return -1;
922 }
923
924 // Reverse a gamma table
cmsReverseToneCurveEx(cmsInt32Number nResultSamples,const cmsToneCurve * InCurve)925 cmsToneCurve* CMSEXPORT cmsReverseToneCurveEx(cmsInt32Number nResultSamples, const cmsToneCurve* InCurve)
926 {
927 cmsToneCurve *out;
928 cmsFloat64Number a = 0, b = 0, y, x1, y1, x2, y2;
929 int i, j;
930 int Ascending;
931
932 _cmsAssert(InCurve != NULL);
933
934 // Try to reverse it analytically whatever possible
935
936 if (InCurve ->nSegments == 1 && InCurve ->Segments[0].Type > 0 &&
937 /* InCurve -> Segments[0].Type <= 5 */
938 GetParametricCurveByType(InCurve ->InterpParams->ContextID, InCurve ->Segments[0].Type, NULL) != NULL) {
939
940 return cmsBuildParametricToneCurve(InCurve ->InterpParams->ContextID,
941 -(InCurve -> Segments[0].Type),
942 InCurve -> Segments[0].Params);
943 }
944
945 // Nope, reverse the table.
946 out = cmsBuildTabulatedToneCurve16(InCurve ->InterpParams->ContextID, nResultSamples, NULL);
947 if (out == NULL)
948 return NULL;
949
950 // We want to know if this is an ascending or descending table
951 Ascending = !cmsIsToneCurveDescending(InCurve);
952
953 // Iterate across Y axis
954 for (i=0; i < nResultSamples; i++) {
955
956 y = (cmsFloat64Number) i * 65535.0 / (nResultSamples - 1);
957
958 // Find interval in which y is within.
959 j = GetInterval(y, InCurve->Table16, InCurve->InterpParams);
960 if (j >= 0) {
961
962
963 // Get limits of interval
964 x1 = InCurve ->Table16[j];
965 x2 = InCurve ->Table16[j+1];
966
967 y1 = (cmsFloat64Number) (j * 65535.0) / (InCurve ->nEntries - 1);
968 y2 = (cmsFloat64Number) ((j+1) * 65535.0 ) / (InCurve ->nEntries - 1);
969
970 // If collapsed, then use any
971 if (x1 == x2) {
972
973 out ->Table16[i] = _cmsQuickSaturateWord(Ascending ? y2 : y1);
974 continue;
975
976 } else {
977
978 // Interpolate
979 a = (y2 - y1) / (x2 - x1);
980 b = y2 - a * x2;
981 }
982 }
983
984 out ->Table16[i] = _cmsQuickSaturateWord(a* y + b);
985 }
986
987
988 return out;
989 }
990
991 // Reverse a gamma table
cmsReverseToneCurve(const cmsToneCurve * InGamma)992 cmsToneCurve* CMSEXPORT cmsReverseToneCurve(const cmsToneCurve* InGamma)
993 {
994 _cmsAssert(InGamma != NULL);
995
996 return cmsReverseToneCurveEx(4096, InGamma);
997 }
998
999 // From: Eilers, P.H.C. (1994) Smoothing and interpolation with finite
1000 // differences. in: Graphic Gems IV, Heckbert, P.S. (ed.), Academic press.
1001 //
1002 // Smoothing and interpolation with second differences.
1003 //
1004 // Input: weights (w), data (y): vector from 1 to m.
1005 // Input: smoothing parameter (lambda), length (m).
1006 // Output: smoothed vector (z): vector from 1 to m.
1007
1008 static
smooth2(cmsContext ContextID,cmsFloat32Number w[],cmsFloat32Number y[],cmsFloat32Number z[],cmsFloat32Number lambda,int m)1009 cmsBool smooth2(cmsContext ContextID, cmsFloat32Number w[], cmsFloat32Number y[], cmsFloat32Number z[], cmsFloat32Number lambda, int m)
1010 {
1011 int i, i1, i2;
1012 cmsFloat32Number *c, *d, *e;
1013 cmsBool st;
1014
1015
1016 c = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1017 d = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1018 e = (cmsFloat32Number*) _cmsCalloc(ContextID, MAX_NODES_IN_CURVE, sizeof(cmsFloat32Number));
1019
1020 if (c != NULL && d != NULL && e != NULL) {
1021
1022
1023 d[1] = w[1] + lambda;
1024 c[1] = -2 * lambda / d[1];
1025 e[1] = lambda /d[1];
1026 z[1] = w[1] * y[1];
1027 d[2] = w[2] + 5 * lambda - d[1] * c[1] * c[1];
1028 c[2] = (-4 * lambda - d[1] * c[1] * e[1]) / d[2];
1029 e[2] = lambda / d[2];
1030 z[2] = w[2] * y[2] - c[1] * z[1];
1031
1032 for (i = 3; i < m - 1; i++) {
1033 i1 = i - 1; i2 = i - 2;
1034 d[i]= w[i] + 6 * lambda - c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1035 c[i] = (-4 * lambda -d[i1] * c[i1] * e[i1])/ d[i];
1036 e[i] = lambda / d[i];
1037 z[i] = w[i] * y[i] - c[i1] * z[i1] - e[i2] * z[i2];
1038 }
1039
1040 i1 = m - 2; i2 = m - 3;
1041
1042 d[m - 1] = w[m - 1] + 5 * lambda -c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1043 c[m - 1] = (-2 * lambda - d[i1] * c[i1] * e[i1]) / d[m - 1];
1044 z[m - 1] = w[m - 1] * y[m - 1] - c[i1] * z[i1] - e[i2] * z[i2];
1045 i1 = m - 1; i2 = m - 2;
1046
1047 d[m] = w[m] + lambda - c[i1] * c[i1] * d[i1] - e[i2] * e[i2] * d[i2];
1048 z[m] = (w[m] * y[m] - c[i1] * z[i1] - e[i2] * z[i2]) / d[m];
1049 z[m - 1] = z[m - 1] / d[m - 1] - c[m - 1] * z[m];
1050
1051 for (i = m - 2; 1<= i; i--)
1052 z[i] = z[i] / d[i] - c[i] * z[i + 1] - e[i] * z[i + 2];
1053
1054 st = TRUE;
1055 }
1056 else st = FALSE;
1057
1058 if (c != NULL) _cmsFree(ContextID, c);
1059 if (d != NULL) _cmsFree(ContextID, d);
1060 if (e != NULL) _cmsFree(ContextID, e);
1061
1062 return st;
1063 }
1064
1065 // Smooths a curve sampled at regular intervals.
cmsSmoothToneCurve(cmsToneCurve * Tab,cmsFloat64Number lambda)1066 cmsBool CMSEXPORT cmsSmoothToneCurve(cmsToneCurve* Tab, cmsFloat64Number lambda)
1067 {
1068 cmsFloat32Number w[MAX_NODES_IN_CURVE], y[MAX_NODES_IN_CURVE], z[MAX_NODES_IN_CURVE];
1069 int i, nItems, Zeros, Poles;
1070
1071 if (Tab == NULL) return FALSE;
1072
1073 if (cmsIsToneCurveLinear(Tab)) return TRUE; // Nothing to do
1074
1075 nItems = Tab -> nEntries;
1076
1077 if (nItems >= MAX_NODES_IN_CURVE) {
1078 cmsSignalError(Tab ->InterpParams->ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: too many points.");
1079 return FALSE;
1080 }
1081
1082 memset(w, 0, nItems * sizeof(cmsFloat32Number));
1083 memset(y, 0, nItems * sizeof(cmsFloat32Number));
1084 memset(z, 0, nItems * sizeof(cmsFloat32Number));
1085
1086 for (i=0; i < nItems; i++)
1087 {
1088 y[i+1] = (cmsFloat32Number) Tab -> Table16[i];
1089 w[i+1] = 1.0;
1090 }
1091
1092 if (!smooth2(Tab ->InterpParams->ContextID, w, y, z, (cmsFloat32Number) lambda, nItems)) return FALSE;
1093
1094 // Do some reality - checking...
1095 Zeros = Poles = 0;
1096 for (i=nItems; i > 1; --i) {
1097
1098 if (z[i] == 0.) Zeros++;
1099 if (z[i] >= 65535.) Poles++;
1100 if (z[i] < z[i-1]) {
1101 cmsSignalError(Tab ->InterpParams->ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Non-Monotonic.");
1102 return FALSE;
1103 }
1104 }
1105
1106 if (Zeros > (nItems / 3)) {
1107 cmsSignalError(Tab ->InterpParams->ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Degenerated, mostly zeros.");
1108 return FALSE;
1109 }
1110 if (Poles > (nItems / 3)) {
1111 cmsSignalError(Tab ->InterpParams->ContextID, cmsERROR_RANGE, "cmsSmoothToneCurve: Degenerated, mostly poles.");
1112 return FALSE;
1113 }
1114
1115 // Seems ok
1116 for (i=0; i < nItems; i++) {
1117
1118 // Clamp to cmsUInt16Number
1119 Tab -> Table16[i] = _cmsQuickSaturateWord(z[i+1]);
1120 }
1121
1122 return TRUE;
1123 }
1124
1125 // Is a table linear? Do not use parametric since we cannot guarantee some weird parameters resulting
1126 // in a linear table. This way assures it is linear in 12 bits, which should be enought in most cases.
cmsIsToneCurveLinear(const cmsToneCurve * Curve)1127 cmsBool CMSEXPORT cmsIsToneCurveLinear(const cmsToneCurve* Curve)
1128 {
1129 cmsUInt32Number i;
1130 int diff;
1131
1132 _cmsAssert(Curve != NULL);
1133
1134 for (i=0; i < Curve ->nEntries; i++) {
1135
1136 diff = abs((int) Curve->Table16[i] - (int) _cmsQuantizeVal(i, Curve ->nEntries));
1137 if (diff > 0x0f)
1138 return FALSE;
1139 }
1140
1141 return TRUE;
1142 }
1143
1144 // Same, but for monotonicity
cmsIsToneCurveMonotonic(const cmsToneCurve * t)1145 cmsBool CMSEXPORT cmsIsToneCurveMonotonic(const cmsToneCurve* t)
1146 {
1147 int n;
1148 int i, last;
1149 cmsBool lDescending;
1150
1151 _cmsAssert(t != NULL);
1152
1153 // Degenerated curves are monotonic? Ok, let's pass them
1154 n = t ->nEntries;
1155 if (n < 2) return TRUE;
1156
1157 // Curve direction
1158 lDescending = cmsIsToneCurveDescending(t);
1159
1160 if (lDescending) {
1161
1162 last = t ->Table16[0];
1163
1164 for (i = 1; i < n; i++) {
1165
1166 if (t ->Table16[i] - last > 2) // We allow some ripple
1167 return FALSE;
1168 else
1169 last = t ->Table16[i];
1170
1171 }
1172 }
1173 else {
1174
1175 last = t ->Table16[n-1];
1176
1177 for (i = n-2; i >= 0; --i) {
1178
1179 if (t ->Table16[i] - last > 2)
1180 return FALSE;
1181 else
1182 last = t ->Table16[i];
1183
1184 }
1185 }
1186
1187 return TRUE;
1188 }
1189
1190 // Same, but for descending tables
cmsIsToneCurveDescending(const cmsToneCurve * t)1191 cmsBool CMSEXPORT cmsIsToneCurveDescending(const cmsToneCurve* t)
1192 {
1193 _cmsAssert(t != NULL);
1194
1195 return t ->Table16[0] > t ->Table16[t ->nEntries-1];
1196 }
1197
1198
1199 // Another info fn: is out gamma table multisegment?
cmsIsToneCurveMultisegment(const cmsToneCurve * t)1200 cmsBool CMSEXPORT cmsIsToneCurveMultisegment(const cmsToneCurve* t)
1201 {
1202 _cmsAssert(t != NULL);
1203
1204 return t -> nSegments > 1;
1205 }
1206
cmsGetToneCurveParametricType(const cmsToneCurve * t)1207 cmsInt32Number CMSEXPORT cmsGetToneCurveParametricType(const cmsToneCurve* t)
1208 {
1209 _cmsAssert(t != NULL);
1210
1211 if (t -> nSegments != 1) return 0;
1212 return t ->Segments[0].Type;
1213 }
1214
1215 // We need accuracy this time
cmsEvalToneCurveFloat(const cmsToneCurve * Curve,cmsFloat32Number v)1216 cmsFloat32Number CMSEXPORT cmsEvalToneCurveFloat(const cmsToneCurve* Curve, cmsFloat32Number v)
1217 {
1218 _cmsAssert(Curve != NULL);
1219
1220 // Check for 16 bits table. If so, this is a limited-precision tone curve
1221 if (Curve ->nSegments == 0) {
1222
1223 cmsUInt16Number In, Out;
1224
1225 In = (cmsUInt16Number) _cmsQuickSaturateWord(v * 65535.0);
1226 Out = cmsEvalToneCurve16(Curve, In);
1227
1228 return (cmsFloat32Number) (Out / 65535.0);
1229 }
1230
1231 return (cmsFloat32Number) EvalSegmentedFn(Curve, v);
1232 }
1233
1234 // We need xput over here
cmsEvalToneCurve16(const cmsToneCurve * Curve,cmsUInt16Number v)1235 cmsUInt16Number CMSEXPORT cmsEvalToneCurve16(const cmsToneCurve* Curve, cmsUInt16Number v)
1236 {
1237 cmsUInt16Number out;
1238
1239 _cmsAssert(Curve != NULL);
1240
1241 Curve ->InterpParams ->Interpolation.Lerp16(&v, &out, Curve ->InterpParams);
1242 return out;
1243 }
1244
1245
1246 // Least squares fitting.
1247 // A mathematical procedure for finding the best-fitting curve to a given set of points by
1248 // minimizing the sum of the squares of the offsets ("the residuals") of the points from the curve.
1249 // The sum of the squares of the offsets is used instead of the offset absolute values because
1250 // this allows the residuals to be treated as a continuous differentiable quantity.
1251 //
1252 // y = f(x) = x ^ g
1253 //
1254 // R = (yi - (xi^g))
1255 // R2 = (yi - (xi^g))2
1256 // SUM R2 = SUM (yi - (xi^g))2
1257 //
1258 // dR2/dg = -2 SUM x^g log(x)(y - x^g)
1259 // solving for dR2/dg = 0
1260 //
1261 // g = 1/n * SUM(log(y) / log(x))
1262
cmsEstimateGamma(const cmsToneCurve * t,cmsFloat64Number Precision)1263 cmsFloat64Number CMSEXPORT cmsEstimateGamma(const cmsToneCurve* t, cmsFloat64Number Precision)
1264 {
1265 cmsFloat64Number gamma, sum, sum2;
1266 cmsFloat64Number n, x, y, Std;
1267 cmsUInt32Number i;
1268
1269 _cmsAssert(t != NULL);
1270
1271 sum = sum2 = n = 0;
1272
1273 // Excluding endpoints
1274 for (i=1; i < (MAX_NODES_IN_CURVE-1); i++) {
1275
1276 x = (cmsFloat64Number) i / (MAX_NODES_IN_CURVE-1);
1277 y = (cmsFloat64Number) cmsEvalToneCurveFloat(t, (cmsFloat32Number) x);
1278
1279 // Avoid 7% on lower part to prevent
1280 // artifacts due to linear ramps
1281
1282 if (y > 0. && y < 1. && x > 0.07) {
1283
1284 gamma = log(y) / log(x);
1285 sum += gamma;
1286 sum2 += gamma * gamma;
1287 n++;
1288 }
1289 }
1290
1291 // Take a look on SD to see if gamma isn't exponential at all
1292 Std = sqrt((n * sum2 - sum * sum) / (n*(n-1)));
1293
1294 if (Std > Precision)
1295 return -1.0;
1296
1297 return (sum / n); // The mean
1298 }
1299