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 || In ->Segments == NULL || In ->Table16 == 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