2 * DSP functions for Indeo Video Interactive codecs (Indeo4 and Indeo5)
4 * Copyright (c) 2009 Maxim Poliakovski
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * @file libavcodec/ivi_dsp.c
25 * DSP functions (inverse transforms, motion compensation, wavelet recompostions)
26 * for Indeo Video Interactive codecs.
31 #include "ivi_common.h"
34 void ff_ivi_recompose53(const IVIPlaneDesc *plane, uint8_t *dst,
35 const int dst_pitch, const int num_bands)
38 int32_t p0, p1, p2, p3, tmp0, tmp1, tmp2;
39 int32_t b0_1, b0_2, b1_1, b1_2, b1_3, b2_1, b2_2, b2_3, b2_4, b2_5, b2_6;
40 int32_t b3_1, b3_2, b3_3, b3_4, b3_5, b3_6, b3_7, b3_8, b3_9;
41 uint32_t pitch, back_pitch;
42 const IDWTELEM *b0_ptr, *b1_ptr, *b2_ptr, *b3_ptr;
44 /* all bands should have the same pitch */
45 pitch = plane->bands[0].pitch;
47 /* pixels at the position "y-1" will be set to pixels at the "y" for the 1st iteration */
50 /* get pointers to the wavelet bands */
51 b0_ptr = plane->bands[0].buf;
52 b1_ptr = plane->bands[1].buf;
53 b2_ptr = plane->bands[2].buf;
54 b3_ptr = plane->bands[3].buf;
56 for (y = 0; y < plane->height; y += 2) {
57 /* load storage variables with values */
64 b1_1 = b1_ptr[back_pitch];
66 b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch];
70 b2_2 = b2_ptr[0]; // b2[x, y ]
71 b2_3 = b2_2; // b2[x+1,y ] = b2[x,y]
72 b2_5 = b2_ptr[pitch]; // b2[x ,y+1]
73 b2_6 = b2_5; // b2[x+1,y+1] = b2[x,y+1]
77 b3_2 = b3_ptr[back_pitch]; // b3[x ,y-1]
78 b3_3 = b3_2; // b3[x+1,y-1] = b3[x ,y-1]
79 b3_5 = b3_ptr[0]; // b3[x ,y ]
80 b3_6 = b3_5; // b3[x+1,y ] = b3[x ,y ]
81 b3_8 = b3_2 - b3_5*6 + b3_ptr[pitch];
85 for (x = 0, indx = 0; x < plane->width; x+=2, indx++) {
86 /* some values calculated in the previous iterations can */
87 /* be reused in the next ones, so do appropriate copying */
88 b2_1 = b2_2; // b2[x-1,y ] = b2[x, y ]
89 b2_2 = b2_3; // b2[x ,y ] = b2[x+1,y ]
90 b2_4 = b2_5; // b2[x-1,y+1] = b2[x ,y+1]
91 b2_5 = b2_6; // b2[x ,y+1] = b2[x+1,y+1]
92 b3_1 = b3_2; // b3[x-1,y-1] = b3[x ,y-1]
93 b3_2 = b3_3; // b3[x ,y-1] = b3[x+1,y-1]
94 b3_4 = b3_5; // b3[x-1,y ] = b3[x ,y ]
95 b3_5 = b3_6; // b3[x ,y ] = b3[x+1,y ]
96 b3_7 = b3_8; // vert_HPF(x-1)
97 b3_8 = b3_9; // vert_HPF(x )
99 p0 = p1 = p2 = p3 = 0;
101 /* process the LL-band by applying LPF both vertically and horizontally */
105 b0_1 = b0_ptr[indx+1];
106 b0_2 = b0_ptr[pitch+indx+1];
111 p2 = (tmp0 + tmp2) << 3;
112 p3 = (tmp1 + tmp2 + b0_2) << 2;
115 /* process the HL-band by applying HPF vertically and LPF horizontally */
119 b1_2 = b1_ptr[indx+1];
120 b1_1 = b1_ptr[back_pitch+indx+1];
122 tmp2 = tmp1 - tmp0*6 + b1_3;
123 b1_3 = b1_1 - b1_2*6 + b1_ptr[pitch+indx+1];
125 p0 += (tmp0 + tmp1) << 3;
126 p1 += (tmp0 + tmp1 + b1_1 + b1_2) << 2;
128 p3 += (tmp2 + b1_3) << 1;
131 /* process the LH-band by applying LPF vertically and HPF horizontally */
133 b2_3 = b2_ptr[indx+1];
134 b2_6 = b2_ptr[pitch+indx+1];
137 tmp1 = b2_1 - b2_2*6 + b2_3;
141 p2 += (tmp0 + b2_4 + b2_5) << 2;
142 p3 += (tmp1 + b2_4 - b2_5*6 + b2_6) << 1;
145 /* process the HH-band by applying HPF both vertically and horizontally */
147 b3_6 = b3_ptr[indx+1]; // b3[x+1,y ]
148 b3_3 = b3_ptr[back_pitch+indx+1]; // b3[x+1,y-1]
154 b3_9 = b3_3 - b3_6*6 + b3_ptr[pitch+indx+1];
156 p0 += (tmp0 + tmp1) << 2;
157 p1 += (tmp0 - tmp1*6 + tmp2) << 1;
158 p2 += (b3_7 + b3_8) << 1;
159 p3 += b3_7 - b3_8*6 + b3_9;
162 /* output four pixels */
163 dst[x] = av_clip_uint8((p0 >> 6) + 128);
164 dst[x+1] = av_clip_uint8((p1 >> 6) + 128);
165 dst[dst_pitch+x] = av_clip_uint8((p2 >> 6) + 128);
166 dst[dst_pitch+x+1] = av_clip_uint8((p3 >> 6) + 128);
169 dst += dst_pitch << 1;
180 /** butterfly operation for the inverse slant transform */
181 #define IVI_SLANT_BFLY(s1, s2, o1, o2, t) \
186 /** This is a reflection a,b = 1/2, 5/4 for the inverse slant transform */
187 #define IVI_IREFLECT(s1, s2, o1, o2, t) \
188 t = ((s1 + s2*2 + 2) >> 2) + s1;\
189 o2 = ((s1*2 - s2 + 2) >> 2) - s2;\
192 /** This is a reflection a,b = 1/2, 7/8 for the inverse slant transform */
193 #define IVI_SLANT_PART4(s1, s2, o1, o2, t) \
194 t = s2 + ((s1*4 - s2 + 4) >> 3);\
195 o2 = s1 + ((-s1 - s2*4 + 4) >> 3);\
198 /** inverse slant8 transform */
199 #define IVI_INV_SLANT8(s1, s4, s8, s5, s2, s6, s3, s7,\
200 d1, d2, d3, d4, d5, d6, d7, d8,\
201 t0, t1, t2, t3, t4, t5, t6, t7, t8) {\
202 IVI_SLANT_PART4(s4, s5, t4, t5, t0);\
204 IVI_SLANT_BFLY(s1, t5, t1, t5, t0); IVI_SLANT_BFLY(s2, s6, t2, t6, t0);\
205 IVI_SLANT_BFLY(s7, s3, t7, t3, t0); IVI_SLANT_BFLY(t4, s8, t4, t8, t0);\
207 IVI_SLANT_BFLY(t1, t2, t1, t2, t0); IVI_IREFLECT (t4, t3, t4, t3, t0);\
208 IVI_SLANT_BFLY(t5, t6, t5, t6, t0); IVI_IREFLECT (t8, t7, t8, t7, t0);\
209 IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
210 IVI_SLANT_BFLY(t5, t8, t5, t8, t0); IVI_SLANT_BFLY(t6, t7, t6, t7, t0);\
211 d1 = COMPENSATE(t1);\
212 d2 = COMPENSATE(t2);\
213 d3 = COMPENSATE(t3);\
214 d4 = COMPENSATE(t4);\
215 d5 = COMPENSATE(t5);\
216 d6 = COMPENSATE(t6);\
217 d7 = COMPENSATE(t7);\
218 d8 = COMPENSATE(t8);}
220 /** inverse slant4 transform */
221 #define IVI_INV_SLANT4(s1, s4, s2, s3, d1, d2, d3, d4, t0, t1, t2, t3, t4) {\
222 IVI_SLANT_BFLY(s1, s2, t1, t2, t0); IVI_IREFLECT (s4, s3, t4, t3, t0);\
224 IVI_SLANT_BFLY(t1, t4, t1, t4, t0); IVI_SLANT_BFLY(t2, t3, t2, t3, t0);\
225 d1 = COMPENSATE(t1);\
226 d2 = COMPENSATE(t2);\
227 d3 = COMPENSATE(t3);\
228 d4 = COMPENSATE(t4);}
230 void ff_ivi_inverse_slant_8x8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
236 int t0, t1, t2, t3, t4, t5, t6, t7, t8;
238 #define COMPENSATE(x) (x)
241 for (i = 0; i < 8; i++) {
243 IVI_INV_SLANT8(src[0], src[8], src[16], src[24], src[32], src[40], src[48], src[56],
244 dst[0], dst[8], dst[16], dst[24], dst[32], dst[40], dst[48], dst[56],
245 t0, t1, t2, t3, t4, t5, t6, t7, t8);
247 dst[0] = dst[8] = dst[16] = dst[24] = dst[32] = dst[40] = dst[48] = dst[56] = 0;
254 #define COMPENSATE(x) ((x + 1)>>1)
256 for (i = 0; i < 8; i++) {
257 if (!src[0] && !src[1] && !src[2] && !src[3] && !src[4] && !src[5] && !src[6] && !src[7]) {
258 memset(out, 0, 8*sizeof(out[0]));
260 IVI_INV_SLANT8(src[0], src[1], src[2], src[3], src[4], src[5], src[6], src[7],
261 out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
262 t0, t1, t2, t3, t4, t5, t6, t7, t8);
270 void ff_ivi_inverse_slant_4x4(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
276 int t0, t1, t2, t3, t4;
278 #define COMPENSATE(x) (x)
281 for (i = 0; i < 4; i++) {
283 IVI_INV_SLANT4(src[0], src[4], src[8], src[12],
284 dst[0], dst[4], dst[8], dst[12],
287 dst[0] = dst[4] = dst[8] = dst[12] = 0;
294 #define COMPENSATE(x) ((x + 1)>>1)
296 for (i = 0; i < 4; i++) {
297 if (!src[0] && !src[1] && !src[2] && !src[3]) {
298 out[0] = out[1] = out[2] = out[3] = 0;
300 IVI_INV_SLANT4(src[0], src[1], src[2], src[3],
301 out[0], out[1], out[2], out[3],
310 void ff_ivi_dc_slant_2d(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
315 dc_coeff = (*in + 1) >> 1;
317 for (y = 0; y < blk_size; out += pitch, y++) {
318 for (x = 0; x < blk_size; x++)
323 void ff_ivi_row_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
326 int t0, t1, t2, t3, t4, t5, t6, t7, t8;
328 #define COMPENSATE(x) ((x + 1)>>1)
329 for (i = 0; i < 8; i++) {
330 if (!in[0] && !in[1] && !in[2] && !in[3] && !in[4] && !in[5] && !in[6] && !in[7]) {
331 memset(out, 0, 8*sizeof(out[0]));
333 IVI_INV_SLANT8( in[0], in[1], in[2], in[3], in[4], in[5], in[6], in[7],
334 out[0], out[1], out[2], out[3], out[4], out[5], out[6], out[7],
335 t0, t1, t2, t3, t4, t5, t6, t7, t8);
343 void ff_ivi_dc_row_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
348 dc_coeff = (*in + 1) >> 1;
350 for (x = 0; x < blk_size; x++)
355 for (y = 1; y < blk_size; out += pitch, y++) {
356 for (x = 0; x < blk_size; x++)
361 void ff_ivi_col_slant8(const int32_t *in, int16_t *out, uint32_t pitch, const uint8_t *flags)
363 int i, row2, row4, row8;
364 int t0, t1, t2, t3, t4, t5, t6, t7, t8;
370 #define COMPENSATE(x) ((x + 1)>>1)
371 for (i = 0; i < 8; i++) {
373 IVI_INV_SLANT8(in[0], in[8], in[16], in[24], in[32], in[40], in[48], in[56],
374 out[0], out[pitch], out[row2], out[row2 + pitch], out[row4],
375 out[row4 + pitch], out[row4 + row2], out[row8 - pitch],
376 t0, t1, t2, t3, t4, t5, t6, t7, t8);
378 out[0] = out[pitch] = out[row2] = out[row2 + pitch] = out[row4] =
379 out[row4 + pitch] = out[row4 + row2] = out[row8 - pitch] = 0;
388 void ff_ivi_dc_col_slant(const int32_t *in, int16_t *out, uint32_t pitch, int blk_size)
393 dc_coeff = (*in + 1) >> 1;
395 for (y = 0; y < blk_size; out += pitch, y++) {
397 for (x = 1; x < blk_size; x++)
402 void ff_ivi_put_pixels_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
403 const uint8_t *flags)
407 for (y = 0; y < 8; out += pitch, in += 8, y++)
408 for (x = 0; x < 8; x++)
412 void ff_ivi_put_dc_pixel_8x8(const int32_t *in, int16_t *out, uint32_t pitch,
418 memset(out + 1, 0, 7*sizeof(out[0]));
421 for (y = 1; y < 8; out += pitch, y++)
422 memset(out, 0, 8*sizeof(out[0]));
425 #define IVI_MC_TEMPLATE(size, suffix, OP) \
426 void ff_ivi_mc_ ## size ##x## size ## suffix (int16_t *buf, const int16_t *ref_buf, \
427 uint32_t pitch, int mc_type) \
430 const int16_t *wptr; \
433 case 0: /* fullpel (no interpolation) */ \
434 for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) { \
435 for (j = 0; j < size; j++) {\
436 OP(buf[j], ref_buf[j]); \
440 case 1: /* horizontal halfpel interpolation */ \
441 for (i = 0; i < size; i++, buf += pitch, ref_buf += pitch) \
442 for (j = 0; j < size; j++) \
443 OP(buf[j], (ref_buf[j] + ref_buf[j+1]) >> 1); \
445 case 2: /* vertical halfpel interpolation */ \
446 wptr = ref_buf + pitch; \
447 for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
448 for (j = 0; j < size; j++) \
449 OP(buf[j], (ref_buf[j] + wptr[j]) >> 1); \
451 case 3: /* vertical and horizontal halfpel interpolation */ \
452 wptr = ref_buf + pitch; \
453 for (i = 0; i < size; i++, buf += pitch, wptr += pitch, ref_buf += pitch) \
454 for (j = 0; j < size; j++) \
455 OP(buf[j], (ref_buf[j] + ref_buf[j+1] + wptr[j] + wptr[j+1]) >> 2); \
460 #define OP_PUT(a, b) (a) = (b)
461 #define OP_ADD(a, b) (a) += (b)
463 IVI_MC_TEMPLATE(8, _no_delta, OP_PUT);
464 IVI_MC_TEMPLATE(8, _delta, OP_ADD);
465 IVI_MC_TEMPLATE(4, _no_delta, OP_PUT);
466 IVI_MC_TEMPLATE(4, _delta, OP_ADD);