src/lib/openjp2/mct.c

   1 /*
   2  * The copyright in this software is being made available under the 2-clauses
   3  * BSD License, included below. This software may be subject to other third
   4  * party and contributor rights, including patent rights, and no such rights
   5  * are granted under this license.
   6  *
   7  * Copyright (c) 2002-2014, Universite catholique de Louvain (UCL), Belgium
   8  * Copyright (c) 2002-2014, Professor Benoit Macq
   9  * Copyright (c) 2001-2003, David Janssens
  10  * Copyright (c) 2002-2003, Yannick Verschueren
  11  * Copyright (c) 2003-2007, Francois-Olivier Devaux
  12  * Copyright (c) 2003-2014, Antonin Descampe
  13  * Copyright (c) 2005, Herve Drolon, FreeImage Team
  14  * Copyright (c) 2008, 2011-2012, Centre National d'Etudes Spatiales (CNES), FR
  15  * Copyright (c) 2012, CS Systemes d'Information, France
  16  * All rights reserved.
  17  *
  18  * Redistribution and use in source and binary forms, with or without
  19  * modification, are permitted provided that the following conditions
  20  * are met:
  21  * 1. Redistributions of source code must retain the above copyright
  22  *    notice, this list of conditions and the following disclaimer.
  23  * 2. Redistributions in binary form must reproduce the above copyright
  24  *    notice, this list of conditions and the following disclaimer in the
  25  *    documentation and/or other materials provided with the distribution.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS `AS IS'
  28  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  31  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 #ifdef __SSE__
  41 #include <xmmintrin.h>
  42 #endif
  43
  44 #include "opj_includes.h"
  45
  46 /* <summary> */
  47 /* This table contains the norms of the basis function of the reversible MCT. */
  48 /* </summary> */
  49 static const OPJ_FLOAT64 opj_mct_norms[3] = { 1.732, .8292, .8292 };
  50
  51 /* <summary> */
  52 /* This table contains the norms of the basis function of the irreversible MCT. */
  53 /* </summary> */
  54 static const OPJ_FLOAT64 opj_mct_norms_real[3] = { 1.732, 1.805, 1.573 };
  55
  56 const OPJ_FLOAT64 * opj_mct_get_mct_norms ()
  57 {
  58         return opj_mct_norms;
  59 }
  60
  61 const OPJ_FLOAT64 * opj_mct_get_mct_norms_real ()
  62 {
  63         return opj_mct_norms_real;
  64 }
  65
  66 /* <summary> */
  67 /* Foward reversible MCT. */
  68 /* </summary> */
  69 void opj_mct_encode(
  70                 OPJ_INT32* restrict c0,
  71                 OPJ_INT32* restrict c1,
  72                 OPJ_INT32* restrict c2,
  73                 OPJ_UINT32 n)
  74 {
  75         OPJ_UINT32 i;
  76         for(i = 0; i < n; ++i) {
  77                 OPJ_INT32 r = c0[i];
  78                 OPJ_INT32 g = c1[i];
  79                 OPJ_INT32 b = c2[i];
  80                 OPJ_INT32 y = (r + (g * 2) + b) >> 2;
  81                 OPJ_INT32 u = b - g;
  82                 OPJ_INT32 v = r - g;
  83                 c0[i] = y;
  84                 c1[i] = u;
  85                 c2[i] = v;
  86         }
  87 }
  88
  89 /* <summary> */
  90 /* Inverse reversible MCT. */
  91 /* </summary> */
  92 void opj_mct_decode(
  93                 OPJ_INT32* restrict c0,
  94                 OPJ_INT32* restrict c1,
  95                 OPJ_INT32* restrict c2,
  96                 OPJ_UINT32 n)
  97 {
  98         OPJ_UINT32 i;
  99         for (i = 0; i < n; ++i) {
 100                 OPJ_INT32 y = c0[i];
 101                 OPJ_INT32 u = c1[i];
 102                 OPJ_INT32 v = c2[i];
 103                 OPJ_INT32 g = y - ((u + v) >> 2);
 104                 OPJ_INT32 r = v + g;
 105                 OPJ_INT32 b = u + g;
 106                 c0[i] = r;
 107                 c1[i] = g;
 108                 c2[i] = b;
 109         }
 110 }
 111
 112 /* <summary> */
 113 /* Get norm of basis function of reversible MCT. */
 114 /* </summary> */
 115 OPJ_FLOAT64 opj_mct_getnorm(OPJ_UINT32 compno) {
 116         return opj_mct_norms[compno];
 117 }
 118
 119 /* <summary> */
 120 /* Foward irreversible MCT. */
 121 /* </summary> */
 122 void opj_mct_encode_real(
 123                 OPJ_INT32* restrict c0,
 124                 OPJ_INT32* restrict c1,
 125                 OPJ_INT32* restrict c2,
 126                 OPJ_UINT32 n)
 127 {
 128         OPJ_UINT32 i;
 129         for(i = 0; i < n; ++i) {
 130                 OPJ_INT32 r = c0[i];
 131                 OPJ_INT32 g = c1[i];
 132                 OPJ_INT32 b = c2[i];
 133                 OPJ_INT32 y =  opj_int_fix_mul(r, 2449) + opj_int_fix_mul(g, 4809) + opj_int_fix_mul(b, 934);
 134                 OPJ_INT32 u = -opj_int_fix_mul(r, 1382) - opj_int_fix_mul(g, 2714) + opj_int_fix_mul(b, 4096);
 135                 OPJ_INT32 v =  opj_int_fix_mul(r, 4096) - opj_int_fix_mul(g, 3430) - opj_int_fix_mul(b, 666);
 136                 c0[i] = y;
 137                 c1[i] = u;
 138                 c2[i] = v;
 139         }
 140 }
 141
 142 /* <summary> */
 143 /* Inverse irreversible MCT. */
 144 /* </summary> */
 145 void opj_mct_decode_real(
 146                 OPJ_FLOAT32* restrict c0,
 147                 OPJ_FLOAT32* restrict c1,
 148                 OPJ_FLOAT32* restrict c2,
 149                 OPJ_UINT32 n)
 150 {
 151         OPJ_UINT32 i;
 152 #ifdef __SSE__
 153         __m128 vrv, vgu, vgv, vbu;
 154         vrv = _mm_set1_ps(1.402f);
 155         vgu = _mm_set1_ps(0.34413f);
 156         vgv = _mm_set1_ps(0.71414f);
 157         vbu = _mm_set1_ps(1.772f);
 158         for (i = 0; i < (n >> 3); ++i) {
 159                 __m128 vy, vu, vv;
 160                 __m128 vr, vg, vb;
 161
 162                 vy = _mm_load_ps(c0);
 163                 vu = _mm_load_ps(c1);
 164                 vv = _mm_load_ps(c2);
 165                 vr = _mm_add_ps(vy, _mm_mul_ps(vv, vrv));
 166                 vg = _mm_sub_ps(_mm_sub_ps(vy, _mm_mul_ps(vu, vgu)), _mm_mul_ps(vv, vgv));
 167                 vb = _mm_add_ps(vy, _mm_mul_ps(vu, vbu));
 168                 _mm_store_ps(c0, vr);
 169                 _mm_store_ps(c1, vg);
 170                 _mm_store_ps(c2, vb);
 171                 c0 += 4;
 172                 c1 += 4;
 173                 c2 += 4;
 174
 175                 vy = _mm_load_ps(c0);
 176                 vu = _mm_load_ps(c1);
 177                 vv = _mm_load_ps(c2);
 178                 vr = _mm_add_ps(vy, _mm_mul_ps(vv, vrv));
 179                 vg = _mm_sub_ps(_mm_sub_ps(vy, _mm_mul_ps(vu, vgu)), _mm_mul_ps(vv, vgv));
 180                 vb = _mm_add_ps(vy, _mm_mul_ps(vu, vbu));
 181                 _mm_store_ps(c0, vr);
 182                 _mm_store_ps(c1, vg);
 183                 _mm_store_ps(c2, vb);
 184                 c0 += 4;
 185                 c1 += 4;
 186                 c2 += 4;
 187         }
 188         n &= 7;
 189 #endif
 190         for(i = 0; i < n; ++i) {
 191                 OPJ_FLOAT32 y = c0[i];
 192                 OPJ_FLOAT32 u = c1[i];
 193                 OPJ_FLOAT32 v = c2[i];
 194                 OPJ_FLOAT32 r = y + (v * 1.402f);
 195                 OPJ_FLOAT32 g = y - (u * 0.34413f) - (v * (0.71414f));
 196                 OPJ_FLOAT32 b = y + (u * 1.772f);
 197                 c0[i] = r;
 198                 c1[i] = g;
 199                 c2[i] = b;
 200         }
 201 }
 202
 203 /* <summary> */
 204 /* Get norm of basis function of irreversible MCT. */
 205 /* </summary> */
 206 OPJ_FLOAT64 opj_mct_getnorm_real(OPJ_UINT32 compno) {
 207         return opj_mct_norms_real[compno];
 208 }
 209
 210
 211 OPJ_BOOL opj_mct_encode_custom(
 212                                            OPJ_BYTE * pCodingdata,
 213                                            OPJ_UINT32 n,
 214                                            OPJ_BYTE ** pData,
 215                                            OPJ_UINT32 pNbComp,
 216                                            OPJ_UINT32 isSigned)
 217 {
 218         OPJ_FLOAT32 * lMct = (OPJ_FLOAT32 *) pCodingdata;
 219         OPJ_UINT32 i;
 220         OPJ_UINT32 j;
 221         OPJ_UINT32 k;
 222         OPJ_UINT32 lNbMatCoeff = pNbComp * pNbComp;
 223         OPJ_INT32 * lCurrentData = 00;
 224         OPJ_INT32 * lCurrentMatrix = 00;
 225         OPJ_INT32 ** lData = (OPJ_INT32 **) pData;
 226         OPJ_UINT32 lMultiplicator = 1 << 13;
 227         OPJ_INT32 * lMctPtr;
 228
 229     OPJ_ARG_NOT_USED(isSigned);
 230
 231         lCurrentData = (OPJ_INT32 *) opj_malloc((pNbComp + lNbMatCoeff) * sizeof(OPJ_INT32));
 232         if (! lCurrentData) {
 233                 return OPJ_FALSE;
 234         }
 235
 236         lCurrentMatrix = lCurrentData + pNbComp;
 237
 238         for (i =0;i<lNbMatCoeff;++i) {
 239                 lCurrentMatrix[i] = (OPJ_INT32) (*(lMct++) * (OPJ_FLOAT32)lMultiplicator);
 240         }
 241
 242         for (i = 0; i < n; ++i)  {
 243                 lMctPtr = lCurrentMatrix;
 244                 for (j=0;j<pNbComp;++j) {
 245                         lCurrentData[j] = (*(lData[j]));
 246                 }
 247
 248                 for (j=0;j<pNbComp;++j) {
 249                         *(lData[j]) = 0;
 250                         for (k=0;k<pNbComp;++k) {
 251                                 *(lData[j]) += opj_int_fix_mul(*lMctPtr, lCurrentData[k]);
 252                                 ++lMctPtr;
 253                         }
 254
 255                         ++lData[j];
 256                 }
 257         }
 258
 259         opj_free(lCurrentData);
 260
 261         return OPJ_TRUE;
 262 }
 263
 264 OPJ_BOOL opj_mct_decode_custom(
 265                                            OPJ_BYTE * pDecodingData,
 266                                            OPJ_UINT32 n,
 267                                            OPJ_BYTE ** pData,
 268                                            OPJ_UINT32 pNbComp,
 269                                            OPJ_UINT32 isSigned)
 270 {
 271         OPJ_FLOAT32 * lMct;
 272         OPJ_UINT32 i;
 273         OPJ_UINT32 j;
 274         OPJ_UINT32 k;
 275
 276         OPJ_FLOAT32 * lCurrentData = 00;
 277         OPJ_FLOAT32 * lCurrentResult = 00;
 278         OPJ_FLOAT32 ** lData = (OPJ_FLOAT32 **) pData;
 279
 280     OPJ_ARG_NOT_USED(isSigned);
 281
 282         lCurrentData = (OPJ_FLOAT32 *) opj_malloc (2 * pNbComp * sizeof(OPJ_FLOAT32));
 283         if (! lCurrentData) {
 284                 return OPJ_FALSE;
 285         }
 286         lCurrentResult = lCurrentData + pNbComp;
 287
 288         for (i = 0; i < n; ++i) {
 289                 lMct = (OPJ_FLOAT32 *) pDecodingData;
 290                 for (j=0;j<pNbComp;++j) {
 291                         lCurrentData[j] = (OPJ_FLOAT32) (*(lData[j]));
 292                 }
 293                 for (j=0;j<pNbComp;++j) {
 294                         lCurrentResult[j] = 0;
 295                         for     (k=0;k<pNbComp;++k)     {
 296                                 lCurrentResult[j] += *(lMct++) * lCurrentData[k];
 297                         }
 298                         *(lData[j]++) = (OPJ_FLOAT32) (lCurrentResult[j]);
 299                 }
 300         }
 301         opj_free(lCurrentData);
 302         return OPJ_TRUE;
 303 }
 304
 305 void opj_calculate_norms(       OPJ_FLOAT64 * pNorms,
 306                                                         OPJ_UINT32 pNbComps,
 307                                                         OPJ_FLOAT32 * pMatrix)
 308 {
 309         OPJ_UINT32 i,j,lIndex;
 310         OPJ_FLOAT32 lCurrentValue;
 311         OPJ_FLOAT64 * lNorms = (OPJ_FLOAT64 *) pNorms;
 312         OPJ_FLOAT32 * lMatrix = (OPJ_FLOAT32 *) pMatrix;
 313
 314         for     (i=0;i<pNbComps;++i) {
 315                 lNorms[i] = 0;
 316                 lIndex = i;
 317
 318                 for     (j=0;j<pNbComps;++j) {
 319                         lCurrentValue = lMatrix[lIndex];
 320                         lIndex += pNbComps;
 321                         lNorms[i] += lCurrentValue * lCurrentValue;
 322                 }
 323                 lNorms[i] = sqrt(lNorms[i]);
 324         }
 325 }