1/*
2 * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation. Oracle designates this
8 * particular file as subject to the "Classpath" exception as provided
9 * by Oracle in the LICENSE file that accompanied this code.
10 *
11 * This code is distributed in the hope that it will be useful, but WITHOUT
12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 * version 2 for more details (a copy is included in the LICENSE file that
15 * accompanied this code).
16 *
17 * You should have received a copy of the GNU General Public License version
18 * 2 along with this work; if not, write to the Free Software Foundation,
19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 *
21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22 * or visit www.oracle.com if you need additional information or have any
23 * questions.
24 */
25
26
27/*
28 * FUNCTION
29 * mlib_ImageConvMxN - image convolution with edge condition
30 *
31 * SYNOPSIS
32 * mlib_status mlib_ImageConvMxN(mlib_image *dst,
33 * const mlib_image *src,
34 * const mlib_s32 *kernel,
35 * mlib_s32 m,
36 * mlib_s32 n,
37 * mlib_s32 dm,
38 * mlib_s32 dn,
39 * mlib_s32 scale,
40 * mlib_s32 cmask,
41 * mlib_edge edge)
42 *
43 * ARGUMENTS
44 * dst Pointer to destination image.
45 * src Pointer to source image.
46 * m Kernel width (m must be not less than 1).
47 * n Kernel height (n must be not less than 1).
48 * dm, dn Position of key element in convolution kernel.
49 * kernel Pointer to convolution kernel.
50 * scale The scaling factor to convert the input integer
51 * coefficients into floating-point coefficients:
52 * floating-point coefficient = integer coefficient * 2^(-scale)
53 * cmask Channel mask to indicate the channels to be convolved.
54 * Each bit of which represents a channel in the image. The
55 * channels corresponded to 1 bits are those to be processed.
56 * edge Type of edge condition.
57 *
58 * DESCRIPTION
59 * 2-D convolution, MxN kernel.
60 *
61 * The center of the source image is mapped to the center of the
62 * destination image.
63 * The unselected channels are not overwritten. If both src and dst have
64 * just one channel, cmask is ignored.
65 *
66 * The edge condition can be one of the following:
67 * MLIB_EDGE_DST_NO_WRITE (default)
68 * MLIB_EDGE_DST_FILL_ZERO
69 * MLIB_EDGE_DST_COPY_SRC
70 * MLIB_EDGE_SRC_EXTEND
71 *
72 * RESTRICTION
73 * The src and the dst must be the same type and have same number
74 * of channels (1, 2, 3, or 4).
75 * m >= 1, n >= 1,
76 * 0 <= dm < m, 0 <= dn < n.
77 * For data type MLIB_BYTE: 16 <= scale <= 31 (to be compatible with VIS version)
78 * For data type MLIB_USHORT: 17 <= scale <= 32 (to be compatible with VIS version)
79 * For data type MLIB_SHORT: 17 <= scale <= 32 (to be compatible with VIS version)
80 * For data type MLIB_INT: scale >= 0
81 */
82
83#include "mlib_image.h"
84#include "mlib_ImageConv.h"
85
86/***************************************************************/
87static void mlib_ImageConvMxNMulAdd_S32(mlib_d64 *dst,
88 const mlib_s32 *src,
89 const mlib_d64 *dkernel,
90 mlib_s32 n,
91 mlib_s32 m,
92 mlib_s32 nch);
93
94static void mlib_ImageConvMxNMedian_S32(mlib_s32 *dst,
95 mlib_d64 *src,
96 mlib_s32 n,
97 mlib_s32 nch);
98
99static void mlib_ImageConvMxNS322S32_ext(mlib_s32 *dst,
100 const mlib_s32 *src,
101 mlib_s32 n,
102 mlib_s32 nch,
103 mlib_s32 dx_l,
104 mlib_s32 dx_r);
105
106/***************************************************************/
107#ifdef MLIB_USE_FTOI_CLAMPING
108
109#define CLAMP_S32(dst, src) \
110 dst = (mlib_s32)(src)
111
112#else
113
114#define CLAMP_S32(dst, src) { \
115 mlib_d64 s0 = (mlib_d64)(src); \
116 if (s0 > (mlib_d64)MLIB_S32_MAX) s0 = (mlib_d64)MLIB_S32_MAX; \
117 if (s0 < (mlib_d64)MLIB_S32_MIN) s0 = (mlib_d64)MLIB_S32_MIN; \
118 dst = (mlib_s32)s0; \
119}
120
121#endif /* MLIB_USE_FTOI_CLAMPING */
122
123/***************************************************************/
124void mlib_ImageConvMxNMulAdd_S32(mlib_d64 *dst,
125 const mlib_s32 *src,
126 const mlib_d64 *dkernel,
127 mlib_s32 n,
128 mlib_s32 m,
129 mlib_s32 nch)
130{
131 mlib_d64 *dst1 = dst + 1;
132 mlib_s32 i, j;
133
134 for (j = 0; j < m; j += 3, src += 3 * nch, dkernel += 3) {
135 const mlib_s32 *src2 = src + 2 * nch;
136 mlib_d64 hval0 = dkernel[0];
137 mlib_d64 hval1 = dkernel[1];
138 mlib_d64 hval2 = dkernel[2];
139 mlib_d64 val0 = src[0];
140 mlib_d64 val1 = src[nch];
141 mlib_d64 dval = dst[0];
142
143 if (j == m - 2) {
144 hval2 = 0.f;
145 }
146 else if (j == m - 1) {
147 hval1 = 0.f;
148 hval2 = 0.f;
149 }
150
151#ifdef __SUNPRO_C
152#pragma pipeloop(0)
153#endif /* __SUNPRO_C */
154 for (i = 0; i < n; i++) {
155 mlib_d64 dval0 = val0 * hval0 + dval;
156 mlib_d64 val2 = src2[i * nch];
157
158 dval = dst1[i];
159 dval0 += val1 * hval1;
160 dval0 += val2 * hval2;
161 val0 = val1;
162 val1 = val2;
163
164 dst[i] = dval0;
165 }
166 }
167}
168
169/***************************************************************/
170void mlib_ImageConvMxNMedian_S32(mlib_s32 *dst,
171 mlib_d64 *src,
172 mlib_s32 n,
173 mlib_s32 nch)
174{
175 mlib_s32 i;
176
177#ifdef __SUNPRO_C
178#pragma pipeloop(0)
179#endif /* __SUNPRO_C */
180 for (i = 0; i < n; i++) {
181 mlib_s32 res;
182
183 CLAMP_S32(res, src[i]);
184 src[i] = 0.5;
185 dst[i * nch] = res;
186 }
187}
188
189/***************************************************************/
190void mlib_ImageConvMxNS322S32_ext(mlib_s32 *dst,
191 const mlib_s32 *src,
192 mlib_s32 n,
193 mlib_s32 nch,
194 mlib_s32 dx_l,
195 mlib_s32 dx_r)
196{
197 mlib_s32 i;
198 mlib_d64 val = src[0];
199
200 for (i = 0; i < dx_l; i++)
201 dst[i] = (mlib_s32) val;
202#ifdef __SUNPRO_C
203#pragma pipeloop(0)
204#endif /* __SUNPRO_C */
205 for (; i < n - dx_r; i++)
206 dst[i] = src[nch * (i - dx_l)];
207 val = dst[n - dx_r - 1];
208 for (; i < n; i++)
209 dst[i] = (mlib_s32) val;
210}
211
212/***************************************************************/
213mlib_status mlib_convMxNext_s32(mlib_image *dst,
214 const mlib_image *src,
215 const mlib_s32 *kernel,
216 mlib_s32 m,
217 mlib_s32 n,
218 mlib_s32 dx_l,
219 mlib_s32 dx_r,
220 mlib_s32 dy_t,
221 mlib_s32 dy_b,
222 mlib_s32 scale,
223 mlib_s32 cmask)
224{
225 mlib_d64 dspace[1024], *dsa = dspace;
226 mlib_d64 akernel[256], *dkernel = akernel, fscale = 1.0;
227 mlib_s32 wid_e = mlib_ImageGetWidth(src);
228 mlib_d64 *dsh, *dsv;
229 mlib_s32 *isa;
230 mlib_s32 *da = mlib_ImageGetData(dst);
231 mlib_s32 *sa = mlib_ImageGetData(src);
232 mlib_s32 dlb = mlib_ImageGetStride(dst) >> 2;
233 mlib_s32 slb = mlib_ImageGetStride(src) >> 2;
234 mlib_s32 dw = mlib_ImageGetWidth(dst);
235 mlib_s32 dh = mlib_ImageGetHeight(dst);
236 mlib_s32 nch = mlib_ImageGetChannels(dst);
237 mlib_s32 i, j, j1, k, mn;
238
239 /* internal buffer */
240
241 if (3 * wid_e + m > 1024) {
242 dsa = mlib_malloc((3 * wid_e + m) * sizeof(mlib_d64));
243
244 if (dsa == NULL)
245 return MLIB_FAILURE;
246 }
247
248 isa = (mlib_s32 *) dsa;
249
250 /* load kernel */
251 mn = m * n;
252
253 if (mn > 256) {
254 dkernel = mlib_malloc(mn * sizeof(mlib_d64));
255
256 if (dkernel == NULL) {
257 if (dsa != dspace) mlib_free(dsa);
258 return MLIB_FAILURE;
259 }
260 }
261
262 while (scale > 30) {
263 fscale /= (1 << 30);
264 scale -= 30;
265 }
266
267 fscale /= (1 << scale);
268
269 for (i = 0; i < mn; i++) {
270 dkernel[i] = ((mlib_s32 *) kernel)[i] * fscale;
271 }
272
273 dsh = dsa + dw + m;
274 dsv = dsh + dw;
275
276 for (i = 0; i < dw; i++) {
277 dsh[i] = 0.5;
278 dsv[i] = 0.5;
279 }
280
281 for (j = 0; j < dh; j++, da += dlb) {
282 for (k = 0; k < nch; k++)
283 if (cmask & (1 << (nch - 1 - k))) {
284 mlib_s32 *sa1 = sa + k;
285 mlib_d64 *dkernel1 = dkernel;
286
287 for (j1 = 0; j1 < n; j1++, dkernel1 += m) {
288 mlib_ImageConvMxNS322S32_ext(isa, sa1, dw + m - 1, nch, dx_l, dx_r);
289 mlib_ImageConvMxNMulAdd_S32(dsh, isa, dkernel1, dw, m, 1);
290
291 if ((j + j1 >= dy_t) && (j + j1 < dh + n - dy_b - 2))
292 sa1 += slb;
293 }
294
295 mlib_ImageConvMxNMedian_S32(da + k, dsh, dw, nch);
296 }
297
298 if ((j >= dy_t) && (j < dh + n - dy_b - 2))
299 sa += slb;
300 }
301
302 if (dkernel != akernel)
303 mlib_free(dkernel);
304 if (dsa != dspace)
305 mlib_free(dsa);
306 return MLIB_SUCCESS;
307}
308
309/***************************************************************/
310