1 /*
2  * hybriderr.c --
3  *
4  *      Procedures dealing with hybrid2 dithering, which is hybrid
5  *      dithering with error propagation among pixels.
6  *
7  */
8 
9 /*
10  * Copyright (c) 1995 The Regents of the University of California.
11  * All rights reserved.
12  *
13  * Permission to use, copy, modify, and distribute this software and its
14  * documentation for any purpose, without fee, and without written agreement is
15  * hereby granted, provided that the above copyright notice and the following
16  * two paragraphs appear in all copies of this software.
17  *
18  * IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR
19  * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT
20  * OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF
21  * CALIFORNIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
22  *
23  * THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,
24  * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
25  * AND FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
26  * ON AN "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATION TO
27  * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
28  */
29 
30 /* This file contains C code to implement an ordered dither in the
31    luminance channel and F-S error diffusion on chrominance.
32 */
33 
34 #include "video.h"
35 #include "proto.h"
36 #include "dither.h"
37 
38 #define DITH_SIZE 16
39 
40 /* Structures used for hybrid dither with errors propogated. */
41 
42 static unsigned char *l_darrays[DITH_SIZE];
43 static unsigned char *l_darrays0, *l_darrays1, *l_darrays2, *l_darrays3;
44 static unsigned char *l_darrays4, *l_darrays5, *l_darrays6, *l_darrays7;
45 static unsigned char *l_darrays8, *l_darrays9, *l_darrays10, *l_darrays11;
46 static unsigned char *l_darrays12, *l_darrays13, *l_darrays14, *l_darrays15;
47 static unsigned char cr_fsarray[256*256][4];
48 static unsigned char cb_fsarray[256*256][4];
49 static unsigned short  c_fserr[256*256][2];
50 
51 
52 /*
53  *--------------------------------------------------------------
54  *
55  *  InitHybridErrorDither--
56  *
57  *	Initializes structures used for hybrid dither algorithm
58  *      with errors propogated on Cr and Cb.
59  *
60  * Results:
61  *      None.
62  *
63  * Side effects:
64  *      None.
65  *
66  *--------------------------------------------------------------
67  */
68 
69 void
InitHybridErrorDither()70 InitHybridErrorDither()
71 {
72   int i, j, k, err_range, threshval;
73   unsigned char *lmark;
74 
75 
76   for (i=0; i<DITH_SIZE; i++) {
77     lmark = l_darrays[i] = (unsigned char *) malloc(256);
78 
79     for (j=0; j<lum_values[0]; j++) {
80       *lmark++ = 0;
81     }
82 
83     for (j=0; j<(LUM_RANGE-1); j++) {
84       err_range = lum_values[j+1] - lum_values[j];
85       threshval = ((i * err_range) / DITH_SIZE)+lum_values[j];
86 
87       for (k=lum_values[j]; k<lum_values[j+1]; k++) {
88 	if (k > threshval) *lmark++ = ((j+1) * (CR_RANGE * CB_RANGE));
89 	else *lmark++ = (j * (CR_RANGE * CB_RANGE));
90       }
91     }
92 
93     for (j=lum_values[LUM_RANGE-1]; j <256; j++) {
94       *lmark++ = (LUM_RANGE-1)*(CR_RANGE * CB_RANGE);
95     }
96   }
97   l_darrays0 = l_darrays[0]; l_darrays8 = l_darrays[8];
98   l_darrays1 = l_darrays[1]; l_darrays9 = l_darrays[9];
99   l_darrays2 = l_darrays[2]; l_darrays10 = l_darrays[10];
100   l_darrays3 = l_darrays[3]; l_darrays11 = l_darrays[11];
101   l_darrays4 = l_darrays[4]; l_darrays12 = l_darrays[12];
102   l_darrays5 = l_darrays[5]; l_darrays13 = l_darrays[13];
103   l_darrays6 = l_darrays[6]; l_darrays14 = l_darrays[14];
104   l_darrays7 = l_darrays[7]; l_darrays15 = l_darrays[15];
105   {
106     int cr1, cr2, cr3, cr4, err1, err2;
107     int cb1, cb2, cb3, cb4, val, nval;
108     int outerr1, outerr2, outerr3, outerr4;
109     int inerr1, inerr2, inerr3, inerr4;
110     unsigned short oe1, oe2, oe3, oe4;
111 
112     for (j=0; j<65536; j+= 256) {
113 
114       inerr1 = (((j & 0xc000) >> 14)*8) - 12;
115       inerr2 = (((j & 0x3000) >> 12)*8) - 12;
116       inerr3 = (((j & 0x0c00) >> 10)*8) - 12;
117       inerr4 = (((j & 0x0300) >> 8) *8) - 12;
118 
119       for (i=0; i<256; i++) {
120 	val = i;
121 
122 	nval = val+inerr1+inerr3;
123 	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
124 	cr1 = ((nval) * CR_RANGE) / 256;
125 	err1 = ((nval) - cr_values[cr1])/2;
126 	err2 = ((nval) - cr_values[cr1]) - err1;
127 
128 	nval = val+err1+inerr2;
129 	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
130 	cr2 = ((nval) * CR_RANGE) / 256;
131 	err1 = ((nval) - cr_values[cr2])/2;
132 	outerr3 = ((nval) - cr_values[cr2])-err1;
133 
134 	nval = val+err2+inerr4;
135 	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
136 	cr3 = ((nval) * CR_RANGE) / 256;
137 	err2 = ((nval) - cr_values[cr3])/2;
138 	outerr1 = ((nval) - cr_values[cr3]) - err2;
139 
140 	nval = val+err1+err2;
141 	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
142 	cr4 = ((nval) * CR_RANGE) / 256;
143 	outerr2 = ((nval) - cr_values[cr4])/2;
144 	outerr4 = ((nval) - cr_values[cr4])-outerr2;
145 
146 	cr_fsarray[i+j][0] = cr1*CB_RANGE;
147 	cr_fsarray[i+j][1] = cr2*CB_RANGE;
148 	cr_fsarray[i+j][2] = cr3*CB_RANGE;
149 	cr_fsarray[i+j][3] = cr4*CB_RANGE;
150 
151 	if (outerr1 < -16) outerr1++;
152 	else if (outerr1 > 15) outerr1--;
153 	if (outerr2 < -16) outerr2++;
154 	else if (outerr2 > 15) outerr2--;
155 	if (outerr3 < -16) outerr3++;
156 	else if (outerr3 > 15) outerr3--;
157 	if (outerr4 < -16) outerr4++;
158 	else if (outerr4 > 15) outerr4--;
159 
160 	oe1 = (outerr1 + 16) / 8;
161 	oe2 = (outerr2 + 16) / 8;
162 	oe3 = (outerr3 + 16) / 8;
163 	oe4 = (outerr4 + 16) / 8;
164 
165 /* This is a debugging check and should be removed if not needed. */
166 	if ((oe1 > 3) || (oe2 > 3) || (oe3 > 3) || (oe4 > 3))
167 	  fprintf(stderr, "OE error!!!!\n");
168 
169 
170 	c_fserr[i+j][0] = ((oe1 << 14) | (oe2 << 12));
171 
172 	c_fserr[i+j][1] = ((oe3 << 10) | (oe4 << 8));
173       }
174 
175       for (i=0; i<256; i++) {
176 	val = i;
177 	nval = val+inerr1+inerr3;
178   	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
179 	cb1 = ((nval) * CB_RANGE) / 256;
180 	err1 = ((nval) - cb_values[cb1])/2;
181 	err2 = ((nval) - cb_values[cb1]) - err1;
182 
183 	nval = val+err1+inerr2;
184   	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
185 	cb2 = ((nval) * CB_RANGE) / 256;
186 	err1 = ((nval) - cb_values[cb2])/2;
187 
188 	nval = val+err2+inerr4;
189   	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
190 	cb3 = ((nval) * CB_RANGE) / 256;
191 	err2 = ((nval) - cb_values[cb3])/2;
192 
193 	nval = val+err1+err2;
194   	if (nval < 0) nval = 0; else if (nval > 255) nval = 255;
195 	cb4 = ((nval) * CB_RANGE) / 256;
196 
197 	cb_fsarray[i+j][0] = cb1;
198 	cb_fsarray[i+j][1] = cb2;
199 	cb_fsarray[i+j][2] = cb3;
200 	cb_fsarray[i+j][3] = cb4;
201       }
202     }
203   }
204 }
205 
206 
207 /*
208  *--------------------------------------------------------------
209  *
210  * HybridErrorDitherImage --
211  *
212  *	Dithers an image using a hybrid ordered/floyd-steinberg dither.
213  *	Assumptions made:
214  *	  1) The color space is allocated y:cr:cb = 8:4:4
215  *	  2) The spatial resolution of y:cr:cb is 4:1:1
216  *      This dither is almost exactly like the dither implemented in the
217  *      file odith.c (i.e. hybrid dithering) except a quantized amount of
218  *      error is propogated between 2x2 pixel areas in Cr and Cb.
219  *
220  * Results:
221  *	None.
222  *
223  * Side effects:
224  *	None.
225  *
226  *--------------------------------------------------------------
227  */
228 void
HybridErrorDitherImage(lum,cr,cb,out,h,w)229 HybridErrorDitherImage (lum, cr, cb, out, h, w)
230     unsigned char *lum;
231     unsigned char *cr;
232     unsigned char *cb;
233     unsigned char *out;
234     int w, h;
235 {
236   unsigned char *l, *r, *b, *o1, *o2;
237   unsigned char *l2;
238   static int *cr_row_errs;
239   static int *cb_row_errs;
240   int *cr_r_err;
241   int *cb_r_err;
242   int cr_c_err;
243   int cb_c_err;
244   unsigned char *cr_fsptr;
245   unsigned char *cb_fsptr;
246   static int first = 1;
247   int cr_code, cb_code;
248 
249   int i, j;
250   int row_advance, row_advance2;
251   int half_row_advance, half_row_advance2;
252 
253   /* If first time called, allocate error arrays. */
254 
255   if (first) {
256     cr_row_errs = (int *) malloc((w+5)*sizeof(int));
257     cb_row_errs = (int *) malloc((w+5)*sizeof(int));
258     first = 0;
259   }
260 
261   row_advance = (w << 1) - 1;
262   row_advance2 = row_advance+2;
263   half_row_advance = (w>>1)-1;
264   half_row_advance2 = half_row_advance+2;
265 
266   l = lum;
267   l2 = lum+w;
268   r = cr;
269   b = cb;
270   o1 = out;
271   o2 = out+w;
272 
273   memset( (char *) cr_row_errs, 0, (w+5)*sizeof(int));
274   cr_r_err = cr_row_errs;
275   cr_c_err = 0;
276   memset( (char *) cb_row_errs, 0, (w+5)*sizeof(int));
277   cb_r_err = cb_row_errs;
278   cb_c_err = 0;
279 
280   for (i=0; i<h; i+=4) {
281 
282     for (j=w; j>0; j-=4) {
283 
284       cr_code = (*cr_r_err | cr_c_err | *r++);
285       cb_code = (*cb_r_err | cb_c_err | *b++);
286 
287       cr_fsptr = cr_fsarray[cr_code];
288       cb_fsptr = cb_fsarray[cb_code];
289 
290       *o1++ = pixel[(l_darrays0[*l++] | *cr_fsptr++ | *cb_fsptr++)];
291       *o1++ = pixel[(l_darrays8[*l++] | *cr_fsptr++ | *cb_fsptr++)];
292       *o2++ = pixel[(l_darrays12[*l2++] | *cr_fsptr++ | *cb_fsptr++)];
293       *o2++ = pixel[(l_darrays4[*l2++] | *cr_fsptr++ | *cb_fsptr++)];
294 
295       cr_c_err = c_fserr[cr_code][1];
296       cb_c_err = c_fserr[cb_code][1];
297       *cr_r_err++ = c_fserr[cr_code][0];
298       *cb_r_err++ = c_fserr[cb_code][0];
299       cr_code = (*cr_r_err | cr_c_err | *r++);
300       cb_code = (*cb_r_err | cb_c_err | *b++);
301 
302       cr_fsptr = cr_fsarray[cr_code];
303       cb_fsptr = cb_fsarray[cb_code];
304 
305       *o1++ = pixel[(l_darrays2[*l++] | *cr_fsptr++ | *cb_fsptr++)];
306       *o1++ = pixel[(l_darrays10[*l++] | *cr_fsptr++ | *cb_fsptr++)];
307       *o2++ = pixel[(l_darrays14[*l2++] | *cr_fsptr++ | *cb_fsptr++)];
308       *o2++ = pixel[(l_darrays6[*l2++] | *cr_fsptr++ | *cb_fsptr++)];
309 
310       cr_c_err = c_fserr[cr_code][1];
311       cb_c_err = c_fserr[cb_code][1];
312       *cr_r_err++ = c_fserr[cr_code][0];
313       *cb_r_err++ = c_fserr[cb_code][0];
314     }
315 
316     l += row_advance; l2 += row_advance;
317     o1 += row_advance; o2 += row_advance;
318     cr_c_err = 0;
319     cb_c_err = 0;
320     cr_r_err--; cb_r_err--;
321     r += half_row_advance; b += half_row_advance;
322 
323     for (j=w; j>0; j-=4) {
324 
325       cr_code = (*cr_r_err | cr_c_err | *r--);
326       cb_code = (*cb_r_err | cb_c_err | *b--);
327       cr_fsptr = cr_fsarray[cr_code];
328       cb_fsptr = cb_fsarray[cb_code];
329 
330       *o1-- = pixel[(l_darrays9[*l--] | *cr_fsptr++ | *cb_fsptr++)];
331       *o1-- = pixel[(l_darrays1[*l--] | *cr_fsptr++ | *cb_fsptr++)];
332       *o2-- = pixel[(l_darrays5[*l2--] | *cr_fsptr++ | *cb_fsptr++)];
333       *o2-- = pixel[(l_darrays13[*l2--] | *cr_fsptr++ | *cb_fsptr++)];
334 
335       cr_c_err = c_fserr[cr_code][1];
336       cb_c_err = c_fserr[cb_code][1];
337       *cr_r_err-- = c_fserr[cr_code][0];
338       *cb_r_err-- = c_fserr[cb_code][0];
339       cr_code = (*cr_r_err | cr_c_err | *r--);
340       cb_code = (*cb_r_err | cb_c_err | *b--);
341       cr_fsptr = cr_fsarray[cr_code];
342       cb_fsptr = cb_fsarray[cb_code];
343 
344       *o1-- = pixel[(l_darrays11[*l--] | *cr_fsptr++ | *cb_fsptr++)];
345       *o1-- = pixel[(l_darrays3[*l--] | *cr_fsptr++ | *cb_fsptr++)];
346       *o2-- = pixel[(l_darrays7[*l2--] | *cr_fsptr++ | *cb_fsptr++)];
347       *o2-- = pixel[(l_darrays15[*l2--] | *cr_fsptr++ | *cb_fsptr++)];
348 
349       cr_c_err = c_fserr[cr_code][1];
350       cb_c_err = c_fserr[cb_code][1];
351       *cr_r_err-- = c_fserr[cr_code][0];
352       *cb_r_err-- = c_fserr[cb_code][0];
353 
354     }
355 
356     l += row_advance2; l2 += row_advance2;
357     o1 += row_advance2; o2 += row_advance2;
358     cr_c_err = 0; cb_c_err = 0;
359     cr_r_err++; cb_r_err++;
360     r += half_row_advance2; b += half_row_advance2;
361   }
362 }
363 
364 
365 
366