custom quant matrices

git-svn-id: svn://svn.videolan.org/x264/trunk@266 df754926-b1dd-0310-bc7b-ec298dee348c

Makefile

@@ -4,7 +4,7 @@ include config.mak
 
 SRCS = common/mc.c common/predict.c common/pixel.c common/macroblock.c \
        common/frame.c common/dct.c common/cpu.c common/cabac.c \
-       common/common.c common/mdate.c common/csp.c \
+       common/common.c common/mdate.c common/csp.c common/set.c\
        encoder/analyse.c encoder/me.c encoder/ratecontrol.c \
        encoder/set.c encoder/macroblock.c encoder/cabac.c \
        encoder/cavlc.c encoder/encoder.c encoder/eval.c

common/common.c

@@ -116,6 +116,14 @@ void    x264_param_default( x264_param_t *param )
     param->analyse.i_chroma_qp_offset = 0;
     param->analyse.b_psnr = 1;
 
+    param->i_cqm_preset = X264_CQM_FLAT;
+    memset( param->cqm_4iy, 16, 16 );
+    memset( param->cqm_4ic, 16, 16 );
+    memset( param->cqm_4py, 16, 16 );
+    memset( param->cqm_4pc, 16, 16 );
+    memset( param->cqm_8iy, 16, 64 );
+    memset( param->cqm_8py, 16, 64 );
+
     param->b_aud = 0;
 }
 

common/common.h

@@ -239,6 +239,11 @@ struct x264_t
     x264_pps_t      *pps;
     int             i_idr_pic_id;
 
+    int             dequant4_mf[4][6][4][4];
+    int             dequant8_mf[2][6][8][8];
+    int             quant4_mf[4][6][4][4];
+    int             quant8_mf[2][6][8][8];
+
     /* Slice header */
     x264_slice_header_t sh;
 

common/macroblock.c

@@ -163,34 +163,35 @@ int x264_mb_transform_8x8_allowed( x264_t *h )
 /****************************************************************************
  * Scan and Quant functions
  ****************************************************************************/
-void x264_mb_dequant_2x2_dc( int16_t dct[2][2], int i_qscale )
+void x264_mb_dequant_2x2_dc( int16_t dct[2][2], int dequant_mf[6][4][4], int i_qscale )
 {
-    const int i_qbits = i_qscale/6 - 1;
+    const int i_qbits = i_qscale/6 - 5;
 
     if( i_qbits >= 0 )
     {
         const int i_dmf = dequant_mf[i_qscale%6][0][0] << i_qbits;
 
-        dct[0][0] = dct[0][0] * i_dmf;
-        dct[0][1] = dct[0][1] * i_dmf;
-        dct[1][0] = dct[1][0] * i_dmf;
-        dct[1][1] = dct[1][1] * i_dmf;
+        dct[0][0] *= i_dmf;
+        dct[0][1] *= i_dmf;
+        dct[1][0] *= i_dmf;
+        dct[1][1] *= i_dmf;
     }
     else
     {
         const int i_dmf = dequant_mf[i_qscale%6][0][0];
+        // chroma DC is truncated, not rounded
 
-        dct[0][0] = ( dct[0][0] * i_dmf ) >> 1;
-        dct[0][1] = ( dct[0][1] * i_dmf ) >> 1;
-        dct[1][0] = ( dct[1][0] * i_dmf ) >> 1;
-        dct[1][1] = ( dct[1][1] * i_dmf ) >> 1;
+        dct[0][0] = ( dct[0][0] * i_dmf ) >> (-i_qbits);
+        dct[0][1] = ( dct[0][1] * i_dmf ) >> (-i_qbits);
+        dct[1][0] = ( dct[1][0] * i_dmf ) >> (-i_qbits);
+        dct[1][1] = ( dct[1][1] * i_dmf ) >> (-i_qbits);
     }
 }
 
-void x264_mb_dequant_4x4_dc( int16_t dct[4][4], int i_qscale )
+void x264_mb_dequant_4x4_dc( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qscale )
 {
-    const int i_qbits = i_qscale/6 - 2;
-    int x,y;
+    const int i_qbits = i_qscale/6 - 6;
+    int y;
 
     if( i_qbits >= 0 )
     {
@@ -198,76 +199,89 @@ void x264_mb_dequant_4x4_dc( int16_t dct[4][4], int i_qscale )
 
         for( y = 0; y < 4; y++ )
         {
-            for( x = 0; x < 4; x++ )
-            {
-                dct[y][x] = dct[y][x] * i_dmf;
-            }
+            dct[y][0] *= i_dmf;
+            dct[y][1] *= i_dmf;
+            dct[y][2] *= i_dmf;
+            dct[y][3] *= i_dmf;
         }
     }
     else
     {
         const int i_dmf = dequant_mf[i_qscale%6][0][0];
-        const int f = -i_qbits; // 1 << (-1-i_qbits)
+        const int f = 1 << (-i_qbits-1);
 
         for( y = 0; y < 4; y++ )
         {
-            for( x = 0; x < 4; x++ )
-            {
-                dct[y][x] = ( dct[y][x] * i_dmf + f ) >> (-i_qbits);
-            }
+            dct[y][0] = ( dct[y][0] * i_dmf + f ) >> (-i_qbits);
+            dct[y][1] = ( dct[y][1] * i_dmf + f ) >> (-i_qbits);
+            dct[y][2] = ( dct[y][2] * i_dmf + f ) >> (-i_qbits);
+            dct[y][3] = ( dct[y][3] * i_dmf + f ) >> (-i_qbits);
         }
     }
 }
 
-void x264_mb_dequant_4x4( int16_t dct[4][4], int i_qscale )
+void x264_mb_dequant_4x4( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qscale )
 {
     const int i_mf = i_qscale%6;
-    const int i_qbits = i_qscale/6;
+    const int i_qbits = i_qscale/6 - 4;
     int y;
 
-    for( y = 0; y < 4; y++ )
+    if( i_qbits >= 0 )
+    {
+        for( y = 0; y < 4; y++ )
+        {
+            dct[y][0] = ( dct[y][0] * dequant_mf[i_mf][y][0] ) << i_qbits;
+            dct[y][1] = ( dct[y][1] * dequant_mf[i_mf][y][1] ) << i_qbits;
+            dct[y][2] = ( dct[y][2] * dequant_mf[i_mf][y][2] ) << i_qbits;
+            dct[y][3] = ( dct[y][3] * dequant_mf[i_mf][y][3] ) << i_qbits;
+        }
+    }
+    else
     {
-        dct[y][0] = ( dct[y][0] * dequant_mf[i_mf][y][0] ) << i_qbits;
-        dct[y][1] = ( dct[y][1] * dequant_mf[i_mf][y][1] ) << i_qbits;
-        dct[y][2] = ( dct[y][2] * dequant_mf[i_mf][y][2] ) << i_qbits;
-        dct[y][3] = ( dct[y][3] * dequant_mf[i_mf][y][3] ) << i_qbits;
+        const int f = 1 << (-i_qbits-1);
+        for( y = 0; y < 4; y++ )
+        {
+            dct[y][0] = ( dct[y][0] * dequant_mf[i_mf][y][0] + f ) >> (-i_qbits);
+            dct[y][1] = ( dct[y][1] * dequant_mf[i_mf][y][1] + f ) >> (-i_qbits);
+            dct[y][2] = ( dct[y][2] * dequant_mf[i_mf][y][2] + f ) >> (-i_qbits);
+            dct[y][3] = ( dct[y][3] * dequant_mf[i_mf][y][3] + f ) >> (-i_qbits);
+        }
     }
 }
 
-void x264_mb_dequant_8x8( int16_t dct[8][8], int i_qscale )
+void x264_mb_dequant_8x8( int16_t dct[8][8], int dequant_mf[6][8][8], int i_qscale )
 {
     const int i_mf = i_qscale%6;
+    const int i_qbits = i_qscale/6 - 6;
     int y;
 
-    if( i_qscale >= 12 )
+    if( i_qbits >= 0 )
     {
-        const int i_qbits = (i_qscale/6) - 2;
         for( y = 0; y < 8; y++ )
         {
-            dct[y][0] = ( dct[y][0] * dequant8_mf[i_mf][y][0] ) << i_qbits;
-            dct[y][1] = ( dct[y][1] * dequant8_mf[i_mf][y][1] ) << i_qbits;
-            dct[y][2] = ( dct[y][2] * dequant8_mf[i_mf][y][2] ) << i_qbits;
-            dct[y][3] = ( dct[y][3] * dequant8_mf[i_mf][y][3] ) << i_qbits;
-            dct[y][4] = ( dct[y][4] * dequant8_mf[i_mf][y][4] ) << i_qbits;
-            dct[y][5] = ( dct[y][5] * dequant8_mf[i_mf][y][5] ) << i_qbits;
-            dct[y][6] = ( dct[y][6] * dequant8_mf[i_mf][y][6] ) << i_qbits;
-            dct[y][7] = ( dct[y][7] * dequant8_mf[i_mf][y][7] ) << i_qbits;
+            dct[y][0] = ( dct[y][0] * dequant_mf[i_mf][y][0] ) << i_qbits;
+            dct[y][1] = ( dct[y][1] * dequant_mf[i_mf][y][1] ) << i_qbits;
+            dct[y][2] = ( dct[y][2] * dequant_mf[i_mf][y][2] ) << i_qbits;
+            dct[y][3] = ( dct[y][3] * dequant_mf[i_mf][y][3] ) << i_qbits;
+            dct[y][4] = ( dct[y][4] * dequant_mf[i_mf][y][4] ) << i_qbits;
+            dct[y][5] = ( dct[y][5] * dequant_mf[i_mf][y][5] ) << i_qbits;
+            dct[y][6] = ( dct[y][6] * dequant_mf[i_mf][y][6] ) << i_qbits;
+            dct[y][7] = ( dct[y][7] * dequant_mf[i_mf][y][7] ) << i_qbits;
         }
     }
     else
     {
-        const int i_qbits = 2 - (i_qscale/6);
-        const int i_round = i_qbits; // 1<<(i_qbits-1)
+        const int f = 1 << (-i_qbits-1);
         for( y = 0; y < 8; y++ )
         {
-            dct[y][0] = ( dct[y][0] * dequant8_mf[i_mf][y][0] + i_round ) >> i_qbits;
-            dct[y][1] = ( dct[y][1] * dequant8_mf[i_mf][y][1] + i_round ) >> i_qbits;
-            dct[y][2] = ( dct[y][2] * dequant8_mf[i_mf][y][2] + i_round ) >> i_qbits;
-            dct[y][3] = ( dct[y][3] * dequant8_mf[i_mf][y][3] + i_round ) >> i_qbits;
-            dct[y][4] = ( dct[y][4] * dequant8_mf[i_mf][y][4] + i_round ) >> i_qbits;
-            dct[y][5] = ( dct[y][5] * dequant8_mf[i_mf][y][5] + i_round ) >> i_qbits;
-            dct[y][6] = ( dct[y][6] * dequant8_mf[i_mf][y][6] + i_round ) >> i_qbits;
-            dct[y][7] = ( dct[y][7] * dequant8_mf[i_mf][y][7] + i_round ) >> i_qbits;
+            dct[y][0] = ( dct[y][0] * dequant_mf[i_mf][y][0] + f ) >> (-i_qbits);
+            dct[y][1] = ( dct[y][1] * dequant_mf[i_mf][y][1] + f ) >> (-i_qbits);
+            dct[y][2] = ( dct[y][2] * dequant_mf[i_mf][y][2] + f ) >> (-i_qbits);
+            dct[y][3] = ( dct[y][3] * dequant_mf[i_mf][y][3] + f ) >> (-i_qbits);
+            dct[y][4] = ( dct[y][4] * dequant_mf[i_mf][y][4] + f ) >> (-i_qbits);
+            dct[y][5] = ( dct[y][5] * dequant_mf[i_mf][y][5] + f ) >> (-i_qbits);
+            dct[y][6] = ( dct[y][6] * dequant_mf[i_mf][y][6] + f ) >> (-i_qbits);
+            dct[y][7] = ( dct[y][7] * dequant_mf[i_mf][y][7] + f ) >> (-i_qbits);
         }
     }
 }

common/macroblock.h

@@ -157,6 +157,19 @@ static const int x264_mb_partition_count_table[17] =
     4, 2, 2, 1
 };
 
+static const int x264_zigzag_scan4[16] =
+{
+    0,  1,  4,  8,  5,  2,  3,  6,  9, 12, 13, 10,  7, 11, 14, 15
+};
+static const int x264_zigzag_scan8[64] =
+{
+    0,  1,  8, 16,  9,  2,  3, 10, 17, 24, 32, 25, 18, 11,  4,  5,
+   12, 19, 26, 33, 40, 48, 41, 34, 27, 20, 13,  6,  7, 14, 21, 28,
+   35, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, 30, 37, 44, 51,
+   58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63
+};  
+
+
 void x264_macroblock_cache_init( x264_t *h );
 void x264_macroblock_slice_init( x264_t *h );
 void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y );
@@ -165,10 +178,10 @@ void x264_macroblock_cache_end( x264_t *h );
 
 void x264_macroblock_bipred_init( x264_t *h );
 
-void x264_mb_dequant_4x4_dc( int16_t dct[4][4], int i_qscale );
-void x264_mb_dequant_2x2_dc( int16_t dct[2][2], int i_qscale );
-void x264_mb_dequant_4x4( int16_t dct[4][4], int i_qscale );
-void x264_mb_dequant_8x8( int16_t dct[8][8], int i_qscale );
+void x264_mb_dequant_4x4_dc( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qscale );
+void x264_mb_dequant_2x2_dc( int16_t dct[2][2], int dequant_mf[6][4][4], int i_qscale );
+void x264_mb_dequant_4x4( int16_t dct[4][4], int dequant_mf[6][4][4], int i_qscale );
+void x264_mb_dequant_8x8( int16_t dct[8][8], int dequant_mf[6][8][8], int i_qscale );
 
 /* x264_mb_predict_mv_16x16:
  *      set mvp with predicted mv for D_16x16 block

common/set.c

+/*****************************************************************************
+ * set.c: h264 encoder library
+ *****************************************************************************
+ * Copyright (C) 2005 x264 project
+ *
+ * Authors: Loren Merritt <lorenm@u.washington.edu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
+ *****************************************************************************/
+
+#include "common.h"
+
+static const int dequant4_scale[6][3] =
+{
+    { 10, 13, 16 },
+    { 11, 14, 18 },
+    { 13, 16, 20 },
+    { 14, 18, 23 },
+    { 16, 20, 25 },
+    { 18, 23, 29 }
+};
+static const int quant4_scale[6][3] =
+{
+    { 13107, 8066, 5243 },
+    { 11916, 7490, 4660 },
+    { 10082, 6554, 4194 },
+    {  9362, 5825, 3647 },
+    {  8192, 5243, 3355 },
+    {  7282, 4559, 2893 },
+};
+
+static const int quant8_scan[16] =
+{
+    0,3,4,3, 3,1,5,1, 4,5,2,5, 3,1,5,1
+};
+static const int dequant8_scale[6][6] =
+{
+    { 20, 18, 32, 19, 25, 24 },
+    { 22, 19, 35, 21, 28, 26 },
+    { 26, 23, 42, 24, 33, 31 },
+    { 28, 25, 45, 26, 35, 33 },
+    { 32, 28, 51, 30, 40, 38 },
+    { 36, 32, 58, 34, 46, 43 },
+};
+static const int quant8_scale[6][6] =
+{
+    { 13107, 11428, 20972, 12222, 16777, 15481 },
+    { 11916, 10826, 19174, 11058, 14980, 14290 },
+    { 10082,  8943, 15978,  9675, 12710, 11985 },
+    {  9362,  8228, 14913,  8931, 11984, 11259 },
+    {  8192,  7346, 13159,  7740, 10486,  9777 },
+    {  7282,  6428, 11570,  6830,  9118,  8640 }
+};
+
+void x264_cqm_init( x264_t *h )
+{
+    int def_quant4[6][16];
+    int def_quant8[6][64];
+    int def_dequant4[6][16];
+    int def_dequant8[6][64];
+    int q, i, i_list;
+
+    for( q = 0; q < 6; q++ )
+    {
+        for( i = 0; i < 16; i++ )
+        {
+            int j = (i&1) + ((i>>2)&1);
+            def_dequant4[q][i] = dequant4_scale[q][j];
+            def_quant4[q][i]   =   quant4_scale[q][j];
+        }
+        for( i = 0; i < 64; i++ )
+        {
+            int j = quant8_scan[((i>>1)&12) | (i&3)];
+            def_dequant8[q][i] = dequant8_scale[q][j];
+            def_quant8[q][i]   =   quant8_scale[q][j];
+        }
+    }
+
+    for( q = 0; q < 6; q++ )
+    {
+        for( i_list = 0; i_list < 4; i_list++ )
+            for( i = 0; i < 16; i++ )
+            {
+                h->dequant4_mf[i_list][q][0][i] = def_dequant4[q][i] * h->pps->scaling_list[i_list][i];
+                h->  quant4_mf[i_list][q][0][i] = def_quant4[q][i] * 16 / h->pps->scaling_list[i_list][i];
+            }
+        for( i_list = 0; i_list < 2; i_list++ )
+            for( i = 0; i < 64; i++ )
+            {
+                h->dequant8_mf[i_list][q][0][i] = def_dequant8[q][i] * h->pps->scaling_list[4+i_list][i];
+                h->  quant8_mf[i_list][q][0][i] = def_quant8[q][i] * 16 / h->pps->scaling_list[4+i_list][i];
+            }
+    }
+}
+

common/set.h

@@ -35,6 +35,19 @@ enum profile_e
     PROFILE_HIGH444 = 144
 };
 
+enum cqm4_e
+{
+    CQM_4IY = 0,
+    CQM_4PY = 1,
+    CQM_4IC = 2,
+    CQM_4PC = 3
+};
+enum cqm8_e
+{
+    CQM_8IY = 0,
+    CQM_8PY = 1
+};
+
 typedef struct
 {
     int i_id;
@@ -147,6 +160,60 @@ typedef struct
 
     int b_transform_8x8_mode;
 
+    int i_cqm_preset;
+    const uint8_t *scaling_list[6]; /* could be 8, but we don't allow separate Cb/Cr lists */
+
 } x264_pps_t;
 
+/* default quant matrices */
+static const uint8_t x264_cqm_jvt4i[16] =
+{
+      6,13,20,28,
+     13,20,28,32,
+     20,28,32,37,
+     28,32,37,42
+};
+static const uint8_t x264_cqm_jvt4p[16] =
+{
+    10,14,20,24,
+    14,20,24,27,
+    20,24,27,30,
+    24,27,30,34
+};
+static const uint8_t x264_cqm_jvt8i[64] =
+{
+     6,10,13,16,18,23,25,27,
+    10,11,16,18,23,25,27,29,
+    13,16,18,23,25,27,29,31,
+    16,18,23,25,27,29,31,33,
+    18,23,25,27,29,31,33,36,
+    23,25,27,29,31,33,36,38,
+    25,27,29,31,33,36,38,40,
+    27,29,31,33,36,38,40,42
+};
+static const uint8_t x264_cqm_jvt8p[64] =
+{
+     9,13,15,17,19,21,22,24,
+    13,13,17,19,21,22,24,25,
+    15,17,19,21,22,24,25,27,
+    17,19,21,22,24,25,27,28,
+    19,21,22,24,25,27,28,30,
+    21,22,24,25,27,28,30,32,
+    22,24,25,27,28,30,32,33,
+    24,25,27,28,30,32,33,35
+};
+static const uint8_t x264_cqm_flat16[64] =
+{
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16,
+    16,16,16,16,16,16,16,16
+};
+
+void x264_cqm_init( x264_t *h );
+
 #endif

encoder/encoder.c

@@ -387,6 +387,7 @@ static int x264_validate_parameters( x264_t *h )
     {
         h->mb.b_lossless = 1;
         h->param.analyse.b_transform_8x8 = 0;
+        h->param.i_cqm_preset = X264_CQM_FLAT;
         h->param.rc.f_ip_factor = 1;
         h->param.rc.f_pb_factor = 1;
         h->param.analyse.b_psnr = 0;
@@ -414,6 +415,9 @@ static int x264_validate_parameters( x264_t *h )
     if( h->param.i_threads > 1 && h->param.i_cabac_init_idc == -1 )
         h->param.i_cabac_init_idc = 0;
 
+    if( h->param.i_cqm_preset < X264_CQM_FLAT || h->param.i_cqm_preset > X264_CQM_CUSTOM )
+        h->param.i_cqm_preset = X264_CQM_FLAT;
+
     if( h->param.analyse.i_me_method < X264_ME_DIA ||
         h->param.analyse.i_me_method > X264_ME_ESA )
         h->param.analyse.i_me_method = X264_ME_HEX;
@@ -523,6 +527,8 @@ x264_t *x264_encoder_open   ( x264_param_t *param )
 
     h->pps = &h->pps_array[0];
     x264_pps_init( h->pps, 0, &h->param, h->sps);
+
+    x264_cqm_init( h );
     
     h->mb.i_mb_count = h->sps->i_mb_width * h->sps->i_mb_height;
 

encoder/macroblock.c

@@ -45,7 +45,10 @@ static const uint8_t block_idx_xy[4][4] =
     { 5, 7, 13, 15 }
 };
 
-static const int quant_mf[6][4][4] =
+/* def_quant4_mf only for probe_skip; actual encoding uses matrices from set.c */
+/* FIXME this seems to make better decisions with cqm=jvt, but could screw up
+ * with general custom matrices. */
+static const int def_quant4_mf[6][4][4] =
 {
     { { 13107, 8066, 13107, 8066 }, { 8066, 5243, 8066, 5243 },
       { 13107, 8066, 13107, 8066 }, { 8066, 5243, 8066, 5243 } },
@@ -61,65 +64,6 @@ static const int quant_mf[6][4][4] =
       {  7282, 4559,  7282, 4559 }, { 4559, 2893, 4559, 2893 } }
 };
 
-const int quant8_mf[6][8][8] =
-{
-  {
-    { 13107, 12222, 16777, 12222, 13107, 12222, 16777, 12222 },
-    { 12222, 11428, 15481, 11428, 12222, 11428, 15481, 11428 },
-    { 16777, 15481, 20972, 15481, 16777, 15481, 20972, 15481 },
-    { 12222, 11428, 15481, 11428, 12222, 11428, 15481, 11428 },
-    { 13107, 12222, 16777, 12222, 13107, 12222, 16777, 12222 },
-    { 12222, 11428, 15481, 11428, 12222, 11428, 15481, 11428 },
-    { 16777, 15481, 20972, 15481, 16777, 15481, 20972, 15481 },
-    { 12222, 11428, 15481, 11428, 12222, 11428, 15481, 11428 }
-  }, {
-    { 11916, 11058, 14980, 11058, 11916, 11058, 14980, 11058 },
-    { 11058, 10826, 14290, 10826, 11058, 10826, 14290, 10826 },
-    { 14980, 14290, 19174, 14290, 14980, 14290, 19174, 14290 },
-    { 11058, 10826, 14290, 10826, 11058, 10826, 14290, 10826 },
-    { 11916, 11058, 14980, 11058, 11916, 11058, 14980, 11058 },
-    { 11058, 10826, 14290, 10826, 11058, 10826, 14290, 10826 },
-    { 14980, 14290, 19174, 14290, 14980, 14290, 19174, 14290 },
-    { 11058, 10826, 14290, 10826, 11058, 10826, 14290, 10826 }
-  }, {
-    { 10082,  9675, 12710,  9675, 10082,  9675, 12710,  9675 },
-    {  9675,  8943, 11985,  8943,  9675,  8943, 11985,  8943 },
-    { 12710, 11985, 15978, 11985, 12710, 11985, 15978, 11985 },
-    {  9675,  8943, 11985,  8943,  9675,  8943, 11985,  8943 },
-    { 10082,  9675, 12710,  9675, 10082,  9675, 12710,  9675 },
-    {  9675,  8943, 11985,  8943,  9675,  8943, 11985,  8943 },
-    { 12710, 11985, 15978, 11985, 12710, 11985, 15978, 11985 },
-    {  9675,  8943, 11985,  8943,  9675,  8943, 11985,  8943 }
-  }, {
-    {  9362,  8931, 11984,  8931,  9362,  8931, 11984,  8931 },
-    {  8931,  8228, 11259,  8228,  8931,  8228, 11259,  8228 },
-    { 11984, 11259, 14913, 11259, 11984, 11259, 14913, 11259 },
-    {  8931,  8228, 11259,  8228,  8931,  8228, 11259,  8228 },
-    {  9362,  8931, 11984,  8931,  9362,  8931, 11984,  8931 },
-    {  8931,  8228, 11259,  8228,  8931,  8228, 11259,  8228 },
-    { 11984, 11259, 14913, 11259, 11984, 11259, 14913, 11259 },
-    {  8931,  8228, 11259,  8228,  8931,  8228, 11259,  8228 }
-  }, {
-    {  8192,  7740, 10486,  7740,  8192,  7740, 10486,  7740 },
-    {  7740,  7346,  9777,  7346,  7740,  7346,  9777,  7346 },
-    { 10486,  9777, 13159,  9777, 10486,  9777, 13159,  9777 },
-    {  7740,  7346,  9777,  7346,  7740,  7346,  9777,  7346 },
-    {  8192,  7740, 10486,  7740,  8192,  7740, 10486,  7740 },
-    {  7740,  7346,  9777,  7346,  7740,  7346,  9777,  7346 },
-    { 10486,  9777, 13159,  9777, 10486,  9777, 13159,  9777 },
-    {  7740,  7346,  9777,  7346,  7740,  7346,  9777,  7346 }
-  }, {
-    {  7282,  6830,  9118,  6830,  7282,  6830,  9118,  6830 },
-    {  6830,  6428,  8640,  6428,  6830,  6428,  8640,  6428 },
-    {  9118,  8640, 11570,  8640,  9118,  8640, 11570,  8640 },
-    {  6830,  6428,  8640,  6428,  6830,  6428,  8640,  6428 },
-    {  7282,  6830,  9118,  6830,  7282,  6830,  9118,  6830 },
-    {  6830,  6428,  8640,  6428,  6830,  6428,  8640,  6428 },
-    {  9118,  8640, 11570,  8640,  9118,  8640, 11570,  8640 },
-    {  6830,  6428,  8640,  6428,  6830,  6428,  8640,  6428 }
-  }
-};
-
 static const int i_chroma_qp_table[52] =
 {
      0,  1,  2,  3,  4,  5,  6,  7,  8,  9,
@@ -200,7 +144,7 @@ static inline void sub_zigzag_4x4( int level[15], const uint8_t *p_src, uint8_t
 }
 #undef ZIG
 
-static void quant_8x8( int16_t dct[8][8], int i_qscale, int b_intra )
+static void quant_8x8( int16_t dct[8][8], int quant_mf[6][8][8], int i_qscale, int b_intra )
 {
     const int i_qbits = 16 + i_qscale / 6;
     const int i_mf = i_qscale % 6;
@@ -212,13 +156,13 @@ static void quant_8x8( int16_t dct[8][8], int i_qscale, int b_intra )
         for( x = 0; x < 8; x++ )
         {
             if( dct[y][x] > 0 )
-                dct[y][x] = ( f + dct[y][x] * quant8_mf[i_mf][y][x] ) >> i_qbits;
+                dct[y][x] = ( f + dct[y][x] * quant_mf[i_mf][y][x] ) >> i_qbits;
             else
-                dct[y][x] = - ( ( f - dct[y][x] * quant8_mf[i_mf][y][x] ) >> i_qbits );
+                dct[y][x] = - ( ( f - dct[y][x] * quant_mf[i_mf][y][x] ) >> i_qbits );
         }
     }
 }
-static void quant_4x4( int16_t dct[4][4], int i_qscale, int b_intra )
+static void quant_4x4( int16_t dct[4][4], int quant_mf[6][4][4], int i_qscale, int b_intra )