• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //  qcms
2 //  Copyright (C) 2009 Mozilla Foundation
3 //
4 // Permission is hereby granted, free of charge, to any person obtaining
5 // a copy of this software and associated documentation files (the "Software"),
6 // to deal in the Software without restriction, including without limitation
7 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 // and/or sell copies of the Software, and to permit persons to whom the Software
9 // is furnished to do so, subject to the following conditions:
10 //
11 // The above copyright notice and this permission notice shall be included in
12 // all copies or substantial portions of the Software.
13 //
14 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
15 // EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
16 // THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
17 // NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
18 // LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
19 // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
20 // WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
21 
22 #define _ISOC99_SOURCE  /* for INFINITY */
23 
24 #include <math.h>
25 #include <assert.h>
26 #include <string.h> //memcpy
27 #include "qcmsint.h"
28 #include "transform_util.h"
29 #include "matrix.h"
30 
31 #if !defined(INFINITY)
32 #define INFINITY HUGE_VAL
33 #endif
34 
35 #define PARAMETRIC_CURVE_TYPE 0x70617261 //'para'
36 
37 /* value must be a value between 0 and 1 */
38 //XXX: is the above a good restriction to have?
lut_interp_linear(double value,uint16_t * table,size_t length)39 float lut_interp_linear(double value, uint16_t *table, size_t length)
40 {
41 	int upper, lower;
42 	value = value * (length - 1); // scale to length of the array
43 	upper = ceil(value);
44 	lower = floor(value);
45 	//XXX: can we be more performant here?
46 	value = table[upper]*(1. - (upper - value)) + table[lower]*(upper - value);
47 	/* scale the value */
48 	return value * (1./65535.);
49 }
50 
51 /* same as above but takes and returns a uint16_t value representing a range from 0..1 */
lut_interp_linear16(uint16_t input_value,uint16_t * table,size_t length)52 uint16_t lut_interp_linear16(uint16_t input_value, uint16_t *table, size_t length)
53 {
54 	/* Start scaling input_value to the length of the array: 65535*(length-1).
55 	 * We'll divide out the 65535 next */
56 	uintptr_t value = (input_value * (length - 1));
57 	uint32_t upper = (value + 65534) / 65535; /* equivalent to ceil(value/65535) */
58 	uint32_t lower = value / 65535;           /* equivalent to floor(value/65535) */
59 	/* interp is the distance from upper to value scaled to 0..65535 */
60 	uint32_t interp = value % 65535;
61 
62 	value = (table[upper]*(interp) + table[lower]*(65535 - interp))/65535; // 0..65535*65535
63 
64 	return value;
65 }
66 
67 /* same as above but takes an input_value from 0..PRECACHE_OUTPUT_MAX
68  * and returns a uint8_t value representing a range from 0..1 */
69 static
lut_interp_linear_precache_output(uint32_t input_value,uint16_t * table,size_t length)70 uint8_t lut_interp_linear_precache_output(uint32_t input_value, uint16_t *table, size_t length)
71 {
72 	/* Start scaling input_value to the length of the array: PRECACHE_OUTPUT_MAX*(length-1).
73 	 * We'll divide out the PRECACHE_OUTPUT_MAX next */
74 	uintptr_t value = (input_value * (length - 1));
75 
76 	/* equivalent to ceil(value/PRECACHE_OUTPUT_MAX) */
77 	uint32_t upper = (value + PRECACHE_OUTPUT_MAX-1) / PRECACHE_OUTPUT_MAX;
78 	/* equivalent to floor(value/PRECACHE_OUTPUT_MAX) */
79 	uint32_t lower = value / PRECACHE_OUTPUT_MAX;
80 	/* interp is the distance from upper to value scaled to 0..PRECACHE_OUTPUT_MAX */
81 	uint32_t interp = value % PRECACHE_OUTPUT_MAX;
82 
83 	/* the table values range from 0..65535 */
84 	value = (table[upper]*(interp) + table[lower]*(PRECACHE_OUTPUT_MAX - interp)); // 0..(65535*PRECACHE_OUTPUT_MAX)
85 
86 	/* round and scale */
87 	value += (PRECACHE_OUTPUT_MAX*65535/255)/2;
88         value /= (PRECACHE_OUTPUT_MAX*65535/255); // scale to 0..255
89 	return value;
90 }
91 
92 /* value must be a value between 0 and 1 */
93 //XXX: is the above a good restriction to have?
lut_interp_linear_float(float value,float * table,size_t length)94 float lut_interp_linear_float(float value, float *table, size_t length)
95 {
96         int upper, lower;
97         value = value * (length - 1);
98         upper = ceil(value);
99         lower = floor(value);
100         //XXX: can we be more performant here?
101         value = table[upper]*(1. - (upper - value)) + table[lower]*(upper - value);
102         /* scale the value */
103         return value;
104 }
105 
106 #if 0
107 /* if we use a different representation i.e. one that goes from 0 to 0x1000 we can be more efficient
108  * because we can avoid the divisions and use a shifting instead */
109 /* same as above but takes and returns a uint16_t value representing a range from 0..1 */
110 uint16_t lut_interp_linear16(uint16_t input_value, uint16_t *table, int length)
111 {
112 	uint32_t value = (input_value * (length - 1));
113 	uint32_t upper = (value + 4095) / 4096; /* equivalent to ceil(value/4096) */
114 	uint32_t lower = value / 4096;           /* equivalent to floor(value/4096) */
115 	uint32_t interp = value % 4096;
116 
117 	value = (table[upper]*(interp) + table[lower]*(4096 - interp))/4096; // 0..4096*4096
118 
119 	return value;
120 }
121 #endif
122 
compute_curve_gamma_table_type1(float gamma_table[256],double gamma)123 void compute_curve_gamma_table_type1(float gamma_table[256], double gamma)
124 {
125 	unsigned int i;
126 	for (i = 0; i < 256; i++) {
127 		gamma_table[i] = pow(i/255., gamma);
128 	}
129 }
130 
compute_curve_gamma_table_type2(float gamma_table[256],uint16_t * table,int length)131 void compute_curve_gamma_table_type2(float gamma_table[256], uint16_t *table, int length)
132 {
133 	unsigned int i;
134 	for (i = 0; i < 256; i++) {
135 		gamma_table[i] = lut_interp_linear(i/255., table, length);
136 	}
137 }
138 
compute_curve_gamma_table_type_parametric(float gamma_table[256],float parameter[7],int count)139 void compute_curve_gamma_table_type_parametric(float gamma_table[256], float parameter[7], int count)
140 {
141         size_t X;
142         float interval;
143         float a, b, c, e, f;
144         float y = parameter[0];
145         if (count == 0) {
146                 a = 1;
147                 b = 0;
148                 c = 0;
149                 e = 0;
150                 f = 0;
151                 interval = -INFINITY;
152         } else if(count == 1) {
153                 a = parameter[1];
154                 b = parameter[2];
155                 c = 0;
156                 e = 0;
157                 f = 0;
158                 interval = -1 * parameter[2] / parameter[1];
159         } else if(count == 2) {
160                 a = parameter[1];
161                 b = parameter[2];
162                 c = 0;
163                 e = parameter[3];
164                 f = parameter[3];
165                 interval = -1 * parameter[2] / parameter[1];
166         } else if(count == 3) {
167                 a = parameter[1];
168                 b = parameter[2];
169                 c = parameter[3];
170                 e = -c;
171                 f = 0;
172                 interval = parameter[4];
173         } else if(count == 4) {
174                 a = parameter[1];
175                 b = parameter[2];
176                 c = parameter[3];
177                 e = parameter[5] - c;
178                 f = parameter[6];
179                 interval = parameter[4];
180         } else {
181                 assert(0 && "invalid parametric function type.");
182                 a = 1;
183                 b = 0;
184                 c = 0;
185                 e = 0;
186                 f = 0;
187                 interval = -INFINITY;
188         }
189         for (X = 0; X < 256; X++) {
190                 if (X >= interval) {
191                         // XXX The equations are not exactly as definied in the spec but are
192                         //     algebraic equivilent.
193                         // TODO Should division by 255 be for the whole expression.
194                         gamma_table[X] = pow(a * X / 255. + b, y) + c + e;
195                 } else {
196                         gamma_table[X] = c * X / 255. + f;
197                 }
198         }
199 }
200 
compute_curve_gamma_table_type0(float gamma_table[256])201 void compute_curve_gamma_table_type0(float gamma_table[256])
202 {
203 	unsigned int i;
204 	for (i = 0; i < 256; i++) {
205 		gamma_table[i] = i/255.;
206 	}
207 }
208 
209 
clamp_float(float a)210 float clamp_float(float a)
211 {
212         if (a > 1.)
213                 return 1.;
214         else if (a < 0)
215                 return 0;
216         else
217                 return a;
218 }
219 
clamp_u8(float v)220 unsigned char clamp_u8(float v)
221 {
222 	if (v > 255.)
223 		return 255;
224 	else if (v < 0)
225 		return 0;
226 	else
227 		return floor(v+.5);
228 }
229 
u8Fixed8Number_to_float(uint16_t x)230 float u8Fixed8Number_to_float(uint16_t x)
231 {
232 	// 0x0000 = 0.
233 	// 0x0100 = 1.
234 	// 0xffff = 255  + 255/256
235 	return x/256.;
236 }
237 
238 /* The SSE2 code uses min & max which let NaNs pass through.
239    We want to try to prevent that here by ensuring that
240    gamma table is within expected values. */
validate_gamma_table(float gamma_table[256])241 void validate_gamma_table(float gamma_table[256])
242 {
243 	int i;
244 	for (i = 0; i < 256; i++) {
245 		// Note: we check that the gamma is not in range
246 		// instead of out of range so that we catch NaNs
247 		if (!(gamma_table[i] >= 0.f && gamma_table[i] <= 1.f)) {
248 			gamma_table[i] = 0.f;
249 		}
250 	}
251 }
252 
build_input_gamma_table(struct curveType * TRC)253 float *build_input_gamma_table(struct curveType *TRC)
254 {
255 	float *gamma_table;
256 
257 	if (!TRC) return NULL;
258 	gamma_table = malloc(sizeof(float)*256);
259 	if (gamma_table) {
260 		if (TRC->type == PARAMETRIC_CURVE_TYPE) {
261 			compute_curve_gamma_table_type_parametric(gamma_table, TRC->parameter, TRC->count);
262 		} else {
263 			if (TRC->count == 0) {
264 				compute_curve_gamma_table_type0(gamma_table);
265 			} else if (TRC->count == 1) {
266 				compute_curve_gamma_table_type1(gamma_table, u8Fixed8Number_to_float(TRC->data[0]));
267 			} else {
268 				compute_curve_gamma_table_type2(gamma_table, TRC->data, TRC->count);
269 			}
270 		}
271 	}
272 
273 	validate_gamma_table(gamma_table);
274 
275 	return gamma_table;
276 }
277 
build_colorant_matrix(qcms_profile * p)278 struct matrix build_colorant_matrix(qcms_profile *p)
279 {
280 	struct matrix result;
281 	result.m[0][0] = s15Fixed16Number_to_float(p->redColorant.X);
282 	result.m[0][1] = s15Fixed16Number_to_float(p->greenColorant.X);
283 	result.m[0][2] = s15Fixed16Number_to_float(p->blueColorant.X);
284 	result.m[1][0] = s15Fixed16Number_to_float(p->redColorant.Y);
285 	result.m[1][1] = s15Fixed16Number_to_float(p->greenColorant.Y);
286 	result.m[1][2] = s15Fixed16Number_to_float(p->blueColorant.Y);
287 	result.m[2][0] = s15Fixed16Number_to_float(p->redColorant.Z);
288 	result.m[2][1] = s15Fixed16Number_to_float(p->greenColorant.Z);
289 	result.m[2][2] = s15Fixed16Number_to_float(p->blueColorant.Z);
290 	result.invalid = false;
291 	return result;
292 }
293 
294 /* The following code is copied nearly directly from lcms.
295  * I think it could be much better. For example, Argyll seems to have better code in
296  * icmTable_lookup_bwd and icmTable_setup_bwd. However, for now this is a quick way
297  * to a working solution and allows for easy comparing with lcms. */
lut_inverse_interp16(uint16_t Value,uint16_t LutTable[],int length)298 uint16_fract_t lut_inverse_interp16(uint16_t Value, uint16_t LutTable[], int length)
299 {
300         int l = 1;
301         int r = 0x10000;
302         int x = 0, res;       // 'int' Give spacing for negative values
303         int NumZeroes, NumPoles;
304         int cell0, cell1;
305         double val2;
306         double y0, y1, x0, x1;
307         double a, b, f;
308 
309         // July/27 2001 - Expanded to handle degenerated curves with an arbitrary
310         // number of elements containing 0 at the begining of the table (Zeroes)
311         // and another arbitrary number of poles (FFFFh) at the end.
312         // First the zero and pole extents are computed, then value is compared.
313 
314         NumZeroes = 0;
315         while (LutTable[NumZeroes] == 0 && NumZeroes < length-1)
316                         NumZeroes++;
317 
318         // There are no zeros at the beginning and we are trying to find a zero, so
319         // return anything. It seems zero would be the less destructive choice
320 	/* I'm not sure that this makes sense, but oh well... */
321         if (NumZeroes == 0 && Value == 0)
322             return 0;
323 
324         NumPoles = 0;
325         while (LutTable[length-1- NumPoles] == 0xFFFF && NumPoles < length-1)
326                         NumPoles++;
327 
328         // Does the curve belong to this case?
329         if (NumZeroes > 1 || NumPoles > 1)
330         {
331                 int a, b;
332 
333                 // Identify if value fall downto 0 or FFFF zone
334                 if (Value == 0) return 0;
335                // if (Value == 0xFFFF) return 0xFFFF;
336 
337                 // else restrict to valid zone
338 
339                 a = ((NumZeroes-1) * 0xFFFF) / (length-1);
340                 b = ((length-1 - NumPoles) * 0xFFFF) / (length-1);
341 
342                 l = a - 1;
343                 r = b + 1;
344         }
345 
346 
347         // Seems not a degenerated case... apply binary search
348 
349         while (r > l) {
350 
351                 x = (l + r) / 2;
352 
353 		res = (int) lut_interp_linear16((uint16_fract_t) (x-1), LutTable, length);
354 
355                 if (res == Value) {
356 
357                     // Found exact match.
358 
359                     return (uint16_fract_t) (x - 1);
360                 }
361 
362                 if (res > Value) r = x - 1;
363                 else l = x + 1;
364         }
365 
366         // Not found, should we interpolate?
367 
368 
369         // Get surrounding nodes
370 
371         val2 = (length-1) * ((double) (x - 1) / 65535.0);
372 
373         cell0 = (int) floor(val2);
374         cell1 = (int) ceil(val2);
375 
376         if (cell0 == cell1) return (uint16_fract_t) x;
377 
378         y0 = LutTable[cell0] ;
379         x0 = (65535.0 * cell0) / (length-1);
380 
381         y1 = LutTable[cell1] ;
382         x1 = (65535.0 * cell1) / (length-1);
383 
384         a = (y1 - y0) / (x1 - x0);
385         b = y0 - a * x0;
386 
387         if (fabs(a) < 0.01) return (uint16_fract_t) x;
388 
389         f = ((Value - b) / a);
390 
391         if (f < 0.0) return (uint16_fract_t) 0;
392         if (f >= 65535.0) return (uint16_fract_t) 0xFFFF;
393 
394         return (uint16_fract_t) floor(f + 0.5);
395 
396 }
397 
398 /*
399  The number of entries needed to invert a lookup table should not
400  necessarily be the same as the original number of entries.  This is
401  especially true of lookup tables that have a small number of entries.
402 
403  For example:
404  Using a table like:
405     {0, 3104, 14263, 34802, 65535}
406  invert_lut will produce an inverse of:
407     {3, 34459, 47529, 56801, 65535}
408  which has an maximum error of about 9855 (pixel difference of ~38.346)
409 
410  For now, we punt the decision of output size to the caller. */
invert_lut(uint16_t * table,int length,size_t out_length)411 static uint16_t *invert_lut(uint16_t *table, int length, size_t out_length)
412 {
413         int i;
414         /* for now we invert the lut by creating a lut of size out_length
415          * and attempting to lookup a value for each entry using lut_inverse_interp16 */
416         uint16_t *output = malloc(sizeof(uint16_t)*out_length);
417         if (!output)
418                 return NULL;
419 
420         for (i = 0; i < out_length; i++) {
421                 double x = ((double) i * 65535.) / (double) (out_length - 1);
422                 uint16_fract_t input = floor(x + .5);
423                 output[i] = lut_inverse_interp16(input, table, length);
424         }
425         return output;
426 }
427 
compute_precache_pow(uint8_t * output,float gamma)428 static void compute_precache_pow(uint8_t *output, float gamma)
429 {
430 	uint32_t v = 0;
431 	for (v = 0; v < PRECACHE_OUTPUT_SIZE; v++) {
432 		//XXX: don't do integer/float conversion... and round?
433 		output[v] = 255. * pow(v/(double)PRECACHE_OUTPUT_MAX, gamma);
434 	}
435 }
436 
compute_precache_lut(uint8_t * output,uint16_t * table,int length)437 void compute_precache_lut(uint8_t *output, uint16_t *table, int length)
438 {
439 	uint32_t v = 0;
440 	for (v = 0; v < PRECACHE_OUTPUT_SIZE; v++) {
441 		output[v] = lut_interp_linear_precache_output(v, table, length);
442 	}
443 }
444 
compute_precache_linear(uint8_t * output)445 void compute_precache_linear(uint8_t *output)
446 {
447 	uint32_t v = 0;
448 	for (v = 0; v < PRECACHE_OUTPUT_SIZE; v++) {
449 		//XXX: round?
450 		output[v] = v / (PRECACHE_OUTPUT_SIZE/256);
451 	}
452 }
453 
compute_precache(struct curveType * trc,uint8_t * output)454 qcms_bool compute_precache(struct curveType *trc, uint8_t *output)
455 {
456 
457         if (trc->type == PARAMETRIC_CURVE_TYPE) {
458                         float gamma_table[256];
459                         uint16_t gamma_table_uint[256];
460                         uint16_t i;
461                         uint16_t *inverted;
462                         int inverted_size = 256;
463 
464                         compute_curve_gamma_table_type_parametric(gamma_table, trc->parameter, trc->count);
465                         for(i = 0; i < 256; i++) {
466                                 gamma_table_uint[i] = (uint16_t)(gamma_table[i] * 65535);
467                         }
468 
469                         //XXX: the choice of a minimum of 256 here is not backed by any theory,
470                         //     measurement or data, howeve r it is what lcms uses.
471                         //     the maximum number we would need is 65535 because that's the
472                         //     accuracy used for computing the pre cache table
473                         if (inverted_size < 256)
474                                 inverted_size = 256;
475 
476                         inverted = invert_lut(gamma_table_uint, 256, inverted_size);
477                         if (!inverted)
478                                 return false;
479                         compute_precache_lut(output, inverted, inverted_size);
480                         free(inverted);
481         } else {
482                 if (trc->count == 0) {
483                         compute_precache_linear(output);
484                 } else if (trc->count == 1) {
485                         compute_precache_pow(output, 1./u8Fixed8Number_to_float(trc->data[0]));
486                 } else {
487                         uint16_t *inverted;
488                         int inverted_size = trc->count;
489                         //XXX: the choice of a minimum of 256 here is not backed by any theory,
490                         //     measurement or data, howeve r it is what lcms uses.
491                         //     the maximum number we would need is 65535 because that's the
492                         //     accuracy used for computing the pre cache table
493                         if (inverted_size < 256)
494                                 inverted_size = 256;
495 
496                         inverted = invert_lut(trc->data, trc->count, inverted_size);
497                         if (!inverted)
498                                 return false;
499                         compute_precache_lut(output, inverted, inverted_size);
500                         free(inverted);
501                 }
502         }
503         return true;
504 }
505 
506 
build_linear_table(int length)507 static uint16_t *build_linear_table(int length)
508 {
509         int i;
510         uint16_t *output = malloc(sizeof(uint16_t)*length);
511         if (!output)
512                 return NULL;
513 
514         for (i = 0; i < length; i++) {
515                 double x = ((double) i * 65535.) / (double) (length - 1);
516                 uint16_fract_t input = floor(x + .5);
517                 output[i] = input;
518         }
519         return output;
520 }
521 
build_pow_table(float gamma,int length)522 static uint16_t *build_pow_table(float gamma, int length)
523 {
524         int i;
525         uint16_t *output = malloc(sizeof(uint16_t)*length);
526         if (!output)
527                 return NULL;
528 
529         for (i = 0; i < length; i++) {
530                 uint16_fract_t result;
531                 double x = ((double) i) / (double) (length - 1);
532                 x = pow(x, gamma);                //XXX turn this conversion into a function
533                 result = floor(x*65535. + .5);
534                 output[i] = result;
535         }
536         return output;
537 }
538 
build_output_lut(struct curveType * trc,uint16_t ** output_gamma_lut,size_t * output_gamma_lut_length)539 void build_output_lut(struct curveType *trc,
540                 uint16_t **output_gamma_lut, size_t *output_gamma_lut_length)
541 {
542         if (trc->type == PARAMETRIC_CURVE_TYPE) {
543                 float gamma_table[256];
544                 uint16_t i;
545                 uint16_t *output = malloc(sizeof(uint16_t)*256);
546 
547                 if (!output) {
548                         *output_gamma_lut = NULL;
549                         return;
550                 }
551 
552                 compute_curve_gamma_table_type_parametric(gamma_table, trc->parameter, trc->count);
553                 *output_gamma_lut_length = 256;
554                 for(i = 0; i < 256; i++) {
555                         output[i] = (uint16_t)(gamma_table[i] * 65535);
556                 }
557                 *output_gamma_lut = output;
558         } else {
559                 if (trc->count == 0) {
560                         *output_gamma_lut = build_linear_table(4096);
561                         *output_gamma_lut_length = 4096;
562                 } else if (trc->count == 1) {
563                         float gamma = 1./u8Fixed8Number_to_float(trc->data[0]);
564                         *output_gamma_lut = build_pow_table(gamma, 4096);
565                         *output_gamma_lut_length = 4096;
566                 } else {
567                         //XXX: the choice of a minimum of 256 here is not backed by any theory,
568                         //     measurement or data, however it is what lcms uses.
569                         *output_gamma_lut_length = trc->count;
570                         if (*output_gamma_lut_length < 256)
571                                 *output_gamma_lut_length = 256;
572 
573                         *output_gamma_lut = invert_lut(trc->data, trc->count, *output_gamma_lut_length);
574                 }
575         }
576 
577 }
578