1 |
/****************************************************************************** |
/***************************************************************************** |
2 |
* * |
* |
3 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder * |
* XVID MPEG-4 VIDEO CODEC |
4 |
* * |
* - MacroBlock transfer and quantization - |
5 |
* XviD is an implementation of a part of one or more MPEG-4 Video tools * |
* |
6 |
* as specified in ISO/IEC 14496-2 standard. Those intending to use this * |
* Copyright(C) 2002-2001 Michael Militzer <isibaar@xvid.org> |
7 |
* software module in hardware or software products are advised that its * |
* 2002-2001 Peter Ross <pross@xvid.org> |
8 |
* use may infringe existing patents or copyrights, and any such use * |
* |
9 |
* would be at such party's own risk. The original developer of this * |
* This program is an implementation of a part of one or more MPEG-4 |
10 |
* software module and his/her company, and subsequent editors and their * |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
11 |
* companies, will have no liability for use of this software or * |
* to use this software module in hardware or software products are |
12 |
* modifications or derivatives thereof. * |
* advised that its use may infringe existing patents or copyrights, and |
13 |
* * |
* any such use would be at such party's own risk. The original |
14 |
* XviD is free software; you can redistribute it and/or modify it * |
* developer of this software module and his/her company, and subsequent |
15 |
* under the terms of the GNU General Public License as published by * |
* editors and their companies, will have no liability for use of this |
16 |
* the Free Software Foundation; either version 2 of the License, or * |
* software or modifications or derivatives thereof. |
17 |
* (at your option) any later version. * |
* |
18 |
* * |
* This program is free software; you can redistribute it and/or modify |
19 |
* XviD is distributed in the hope that it will be useful, but * |
* it under the terms of the GNU General Public License as published by |
20 |
* WITHOUT ANY WARRANTY; without even the implied warranty of * |
* the Free Software Foundation; either version 2 of the License, or |
21 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * |
* (at your option) any later version. |
22 |
* GNU General Public License for more details. * |
* |
23 |
* * |
* This program is distributed in the hope that it will be useful, |
24 |
* You should have received a copy of the GNU General Public License * |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
25 |
* along with this program; if not, write to the Free Software * |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
26 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * |
* GNU General Public License for more details. |
27 |
* * |
* |
28 |
******************************************************************************/ |
* You should have received a copy of the GNU General Public License |
29 |
|
* along with this program; if not, write to the Free Software |
30 |
/****************************************************************************** |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
31 |
* * |
* |
32 |
* mbtransquant.c * |
* $Id$ |
33 |
* * |
* |
34 |
* Copyright (C) 2001 - Peter Ross <pross@cs.rmit.edu.au> * |
****************************************************************************/ |
|
* Copyright (C) 2001 - Michael Militzer <isibaar@xvid.org> * |
|
|
* * |
|
|
* For more information visit the XviD homepage: http://www.xvid.org * |
|
|
* * |
|
|
******************************************************************************/ |
|
|
|
|
|
/****************************************************************************** |
|
|
* * |
|
|
* Revision history: * |
|
|
* * |
|
|
* 29.03.2002 interlacing speedup - used transfer strides instead of * |
|
|
* manual field-to-frame conversion * |
|
|
* 26.03.2002 interlacing support - moved transfers outside loops * |
|
|
* 22.12.2001 get_dc_scaler() moved to common.h * |
|
|
* 19.11.2001 introduced coefficient thresholding (Isibaar) * |
|
|
* 17.11.2001 initial version * |
|
|
* * |
|
|
******************************************************************************/ |
|
35 |
|
|
36 |
#include <string.h> |
#include <string.h> |
37 |
|
|
47 |
#include "../quant/quant_h263.h" |
#include "../quant/quant_h263.h" |
48 |
#include "../encoder.h" |
#include "../encoder.h" |
49 |
|
|
50 |
#include "../image/reduced.h" |
#define MIN(X, Y) ((X)<(Y)?(X):(Y)) |
51 |
|
#define MAX(X, Y) ((X)>(Y)?(X):(Y)) |
52 |
|
|
53 |
MBFIELDTEST_PTR MBFieldTest; |
#define TOOSMALL_LIMIT 3 /* skip blocks having a coefficient sum below this value */ |
54 |
|
|
55 |
#define TOOSMALL_LIMIT 1 /* skip blocks having a coefficient sum below this value */ |
/* this isnt pretty, but its better than 20 ifdefs */ |
56 |
|
|
57 |
void |
void |
58 |
MBTransQuantIntra(const MBParam * pParam, |
MBTransQuantIntra(const MBParam * pParam, |
66 |
|
|
67 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
68 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
69 |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
uint32_t next_block = stride * 8; |
70 |
uint32_t i; |
uint32_t i; |
71 |
uint32_t iQuant = frame->quant; |
uint32_t iQuant = frame->quant; |
72 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
73 |
IMAGE *pCurrent = &frame->image; |
IMAGE *pCurrent = &frame->image; |
74 |
|
|
|
start_timer(); |
|
|
if ((frame->global_flags & XVID_REDUCED)) |
|
|
{ |
|
|
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
|
|
|
filter_18x18_to_8x8(&data[0 * 64], pY_Cur, stride); |
|
|
filter_18x18_to_8x8(&data[1 * 64], pY_Cur + 16, stride); |
|
|
filter_18x18_to_8x8(&data[2 * 64], pY_Cur + next_block, stride); |
|
|
filter_18x18_to_8x8(&data[3 * 64], pY_Cur + next_block + 16, stride); |
|
|
filter_18x18_to_8x8(&data[4 * 64], pU_Cur, stride2); |
|
|
filter_18x18_to_8x8(&data[5 * 64], pV_Cur, stride2); |
|
|
}else{ |
|
75 |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
76 |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
77 |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
78 |
|
|
79 |
|
start_timer(); |
80 |
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
transfer_8to16copy(&data[0 * 64], pY_Cur, stride); |
81 |
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
transfer_8to16copy(&data[1 * 64], pY_Cur + 8, stride); |
82 |
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
transfer_8to16copy(&data[2 * 64], pY_Cur + next_block, stride); |
83 |
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
transfer_8to16copy(&data[3 * 64], pY_Cur + next_block + 8, stride); |
84 |
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
transfer_8to16copy(&data[4 * 64], pU_Cur, stride2); |
85 |
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
transfer_8to16copy(&data[5 * 64], pV_Cur, stride2); |
|
} |
|
86 |
stop_transfer_timer(); |
stop_transfer_timer(); |
87 |
|
|
|
/* XXX: rrv+interlacing is buggy */ |
|
88 |
start_timer(); |
start_timer(); |
89 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
90 |
if ((frame->global_flags & XVID_INTERLACING) && |
if ((frame->global_flags & XVID_INTERLACING) && |
105 |
start_timer(); |
start_timer(); |
106 |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
quant_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
107 |
stop_quant_timer(); |
stop_quant_timer(); |
|
} else { |
|
|
start_timer(); |
|
|
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
|
|
stop_quant_timer(); |
|
|
} |
|
108 |
|
|
|
/* speedup: dont decode when encoding only ivops */ |
|
|
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
|
|
{ |
|
|
if (pParam->m_quant_type == H263_QUANT) { |
|
109 |
start_timer(); |
start_timer(); |
110 |
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
dequant_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
111 |
stop_iquant_timer(); |
stop_iquant_timer(); |
112 |
} else { |
} else { |
113 |
start_timer(); |
start_timer(); |
114 |
|
quant4_intra(&qcoeff[i * 64], &data[i * 64], iQuant, iDcScaler); |
115 |
|
stop_quant_timer(); |
116 |
|
|
117 |
|
start_timer(); |
118 |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
dequant4_intra(&data[i * 64], &qcoeff[i * 64], iQuant, iDcScaler); |
119 |
stop_iquant_timer(); |
stop_iquant_timer(); |
120 |
} |
} |
123 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
124 |
stop_idct_timer(); |
stop_idct_timer(); |
125 |
} |
} |
|
} |
|
|
|
|
|
/* speedup: dont decode when encoding only ivops */ |
|
|
if (pParam->iMaxKeyInterval != 1 || pParam->max_bframes > 0) |
|
|
{ |
|
126 |
|
|
127 |
if (pMB->field_dct) { |
if (pMB->field_dct) { |
128 |
next_block = stride; |
next_block = stride; |
130 |
} |
} |
131 |
|
|
132 |
start_timer(); |
start_timer(); |
|
if ((frame->global_flags & XVID_REDUCED)) |
|
|
{ |
|
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block + 16, &data[3 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
|
|
|
}else{ |
|
133 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
134 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
135 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
136 |
transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block + 8, &data[3 * 64], stride); |
137 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
138 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
139 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
} |
|
140 |
|
|
141 |
} |
} |
142 |
|
|
153 |
|
|
154 |
uint32_t stride = pParam->edged_width; |
uint32_t stride = pParam->edged_width; |
155 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
156 |
uint32_t next_block = stride * ((frame->global_flags & XVID_REDUCED)?16:8); |
uint32_t next_block = stride * 8; |
157 |
uint32_t i; |
uint32_t i; |
158 |
uint32_t iQuant = frame->quant; |
uint32_t iQuant = frame->quant; |
159 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
161 |
uint32_t sum; |
uint32_t sum; |
162 |
IMAGE *pCurrent = &frame->image; |
IMAGE *pCurrent = &frame->image; |
163 |
|
|
|
if ((frame->global_flags & XVID_REDUCED)) |
|
|
{ |
|
|
pY_Cur = pCurrent->y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = pCurrent->u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = pCurrent->v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
}else{ |
|
164 |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = pCurrent->y + (y_pos << 4) * stride + (x_pos << 4); |
165 |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * stride2 + (x_pos << 3); |
166 |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
167 |
|
|
168 |
start_timer(); |
start_timer(); |
169 |
pMB->field_dct = 0; |
pMB->field_dct = 0; |
175 |
stop_interlacing_timer(); |
stop_interlacing_timer(); |
176 |
|
|
177 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
uint32_t increase_limit = (iQuant == 1) ? 1 : 0; |
|
|
|
|
178 |
/* |
/* |
179 |
* no need to transfer 8->16-bit |
* no need to transfer 8->16-bit |
180 |
* (this is performed already in motion compensation) |
* (this is performed already in motion compensation) |
193 |
stop_quant_timer(); |
stop_quant_timer(); |
194 |
} |
} |
195 |
|
|
196 |
if ((sum >= TOOSMALL_LIMIT + increase_limit) || (qcoeff[i*64] != 0) || |
if ((sum >= TOOSMALL_LIMIT) || (qcoeff[i*64] != 0) || |
197 |
(qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { |
(qcoeff[i*64+1] != 0) || (qcoeff[i*64+8] != 0)) { |
198 |
|
|
199 |
if (pParam->m_quant_type == H263_QUANT) { |
if (pParam->m_quant_type == H263_QUANT) { |
220 |
} |
} |
221 |
|
|
222 |
start_timer(); |
start_timer(); |
|
if ((frame->global_flags & XVID_REDUCED)) |
|
|
{ |
|
|
if (cbp & 32) |
|
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
}else{ |
|
223 |
if (cbp & 32) |
if (cbp & 32) |
224 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
225 |
if (cbp & 16) |
if (cbp & 16) |
232 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
233 |
if (cbp & 1) |
if (cbp & 1) |
234 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
235 |
stop_transfer_timer(); |
stop_transfer_timer(); |
236 |
|
|
237 |
return cbp; |
return cbp; |
364 |
MBQuantIntra(const MBParam * pParam, |
MBQuantIntra(const MBParam * pParam, |
365 |
FRAMEINFO * frame, |
FRAMEINFO * frame, |
366 |
MACROBLOCK *pMB, |
MACROBLOCK *pMB, |
367 |
int16_t qcoeff[6 * 64], |
int16_t data[6 * 64], |
368 |
int16_t data[6*64]) |
int16_t qcoeff[6 * 64]) |
369 |
{ |
{ |
370 |
int i; |
int i; |
371 |
int iQuant = frame->quant; |
int iQuant = frame->quant; |
558 |
|
|
559 |
|
|
560 |
|
|
561 |
/* permute block and return field dct choice */ |
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
562 |
|
|
563 |
|
|
564 |
uint32_t |
uint32_t |
565 |
MBDecideFieldDCT(int16_t data[6 * 64]) |
MBDecideFieldDCT(int16_t data[6 * 64]) |
566 |
{ |
{ |
|
uint32_t field = MBFieldTest(data); |
|
|
|
|
|
if (field) { |
|
|
MBFrameToField(data); |
|
|
} |
|
|
|
|
|
return field; |
|
|
} |
|
|
|
|
567 |
|
|
|
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
|
|
|
|
|
uint32_t |
|
|
MBFieldTest_c(int16_t data[6 * 64]) |
|
|
{ |
|
568 |
const uint8_t blocks[] = |
const uint8_t blocks[] = |
569 |
{ 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; |
{ 0 * 64, 0 * 64, 0 * 64, 0 * 64, 2 * 64, 2 * 64, 2 * 64, 2 * 64 }; |
570 |
const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; |
const uint8_t lines[] = { 0, 16, 32, 48, 0, 16, 32, 48 }; |
598 |
} |
} |
599 |
} |
} |
600 |
|
|
601 |
return (frame >= (field + 350)); |
if (frame > (field + 350)) { |
602 |
|
MBFrameToField(data); |
603 |
|
} |
604 |
|
|
605 |
|
return (frame > (field + 350)); |
606 |
} |
} |
607 |
|
|
608 |
|
|