shithub: opus

Info  •  Files  •  Log  •  Branches

Add offset and quantization to DRED bitstream

Brings the DRED implementation in line with the -00 draft

--- a/silk/dred_coding.c

+++ b/silk/dred_coding.c

@@ -37,10 +37,20 @@

 #include "celt/cwrs.h"

 #include "celt/laplace.h"

 #include "os_support.h"

+#include "dred_config.h"

 #define LATENT_DIM 80

 #define PVQ_DIM 24

 #define PVQ_K 82

+

+int compute_quantizer(int q0, int dQ, int i) {

+  int quant;

+  static const int dQ_table[8] = {0, 2, 3, 4, 6, 8, 12, 16};

+  quant = 0;

+  quant = q0 + (dQ_table[dQ]*i + 8)/16;

+  return quant > 15 ? 15 : quant;

+  return (int) floor(0.5f + DRED_ENC_Q0 + 1.f * (DRED_ENC_Q1 - DRED_ENC_Q0) * i / (DRED_NUM_REDUNDANCY_FRAMES - 2));

+}

 static void encode_pvq(const int *iy, int N, int K, ec_enc *enc) {

     int fits;

--- a/silk/dred_coding.h

+++ b/silk/dred_coding.h

@@ -31,6 +31,8 @@

 #include "opus_types.h"

 #include "entcode.h"

+int compute_quantizer(int q0, int dQ, int i);

+

 void dred_encode_state(ec_enc *enc, float *x);

 void dred_encode_latents(ec_enc *enc, const float *x, const opus_uint16 *scale, const opus_uint16 *dzone, const opus_uint16 *r, const opus_uint16 *p0);

--- a/silk/dred_config.h

+++ b/silk/dred_config.h

@@ -28,7 +28,7 @@

 #ifndef DRED_CONFIG_H

 #define DRED_CONFIG_H

-#define DRED_VERSION 0

+#define DRED_VERSION 1

 #define DRED_MIN_BYTES 16

 /* these are inpart duplicates to the values defined in dred_rdovae_constants.h */

--- a/silk/dred_decoder.c

+++ b/silk/dred_decoder.c

@@ -48,6 +48,8 @@

   int q_level;

   int i;

   int offset;

+  int q0;

+  int dQ;

   /* since features are decoded in quadruples, it makes no sense to go with an uneven number of redundancy frames */

@@ -55,6 +57,11 @@

   /* decode initial state and initialize RDOVAE decoder */

   ec_dec_init(&ec, (unsigned char*)bytes, num_bytes);

+  dec->dred_offset = ec_dec_uint(&ec, 32);

+  q0 = ec_dec_uint(&ec, 16);

+  dQ = ec_dec_uint(&ec, 8);

+  /*printf("%d %d %d\n", dred_offset, q0, dQ);*/

+

   dred_decode_state(&ec, dec->state);

   /* decode newest to oldest and store oldest to newest */

@@ -63,7 +70,7 @@

       /* FIXME: Figure out how to avoid missing a last frame that would take up < 8 bits. */

       if (8*num_bytes - ec_tell(&ec) <= 7)

          break;

-      q_level = (int) floor(.5 + DRED_ENC_Q0 + 1.f * (DRED_ENC_Q1 - DRED_ENC_Q0) * i / (DRED_NUM_REDUNDANCY_FRAMES - 2));

+      q_level = compute_quantizer(q0, dQ, i/2);

       offset = q_level * DRED_LATENT_DIM;

       dred_decode_latents(

           &ec,

--- a/silk/dred_decoder.h

+++ b/silk/dred_decoder.h

@@ -39,6 +39,7 @@

     float        latents[(DRED_NUM_REDUNDANCY_FRAMES/2)*DRED_LATENT_DIM];

     int          nb_latents;

     int          process_stage;

+    int          dred_offset;

 };

--- a/silk/dred_encoder.c

+++ b/silk/dred_encoder.c

@@ -85,9 +85,18 @@

     int i;

     int offset;

     int ec_buffer_fill;

+    int dred_offset;

+    int q0;

+    int dQ;

     /* entropy coding of state and latents */

     ec_enc_init(&ec_encoder, buf, max_bytes);

+    dred_offset = 8; /* 20 ms */

+    q0 = DRED_ENC_Q0;

+    dQ = 3;

+    ec_enc_uint(&ec_encoder, dred_offset, 32);

+    ec_enc_uint(&ec_encoder, q0, 16);

+    ec_enc_uint(&ec_encoder, dQ, 8);

     dred_encode_state(&ec_encoder, enc->state_buffer);

     for (i = 0; i < IMIN(2*max_chunks, enc->latents_buffer_fill-1); i += 2)

@@ -95,7 +104,7 @@

         ec_enc ec_bak;

         ec_bak = ec_encoder;

-        q_level = (int) floor(0.5f + DRED_ENC_Q0 + 1.f * (DRED_ENC_Q1 - DRED_ENC_Q0) * i / (DRED_NUM_REDUNDANCY_FRAMES - 2));

+        q_level = compute_quantizer(q0, dQ, i/2);

         offset = q_level * DRED_LATENT_DIM;

         dred_encode_latents(

-- 

⑨