[FFmpeg-cvslog] AAC encoder: make pe.min a local minimum

[Claudio Freire]

ffmpeg | branch: master | Claudio Freire <klaussfreire at gmail.com> | Sun Nov 29 15:03:45 2015 -0300| [88e498a87eb255eca48c40fd5570a42372491f2f] | committer: Claudio Freire

AAC encoder: make pe.min a local minimum

As noted in a comment, pe.min in the reference encoder
is centered around current pe. The bit reservoir algo
needs pe.min to be a local minimum, because it can only
account for local PE variations. If it's set to a global
minimum as was being done, bit reservoir logic doesn't
work as efficiently.

This patch tries to forget old minimums and converge to
a local minimum without losing the stability of the
previous solution. Listening tests until now suggest this
solves numerous RC issues.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=88e498a87eb255eca48c40fd5570a42372491f2f
---

 libavcodec/aacpsy.c |   13 ++++++++++---
 tests/fate/aac.mak  |    4 ++--
 2 files changed, 12 insertions(+), 5 deletions(-)

diff --git a/libavcodec/aacpsy.c b/libavcodec/aacpsy.c
index 242df68..40b3b41 100644
--- a/libavcodec/aacpsy.c
+++ b/libavcodec/aacpsy.c
@@ -80,6 +80,8 @@
 #define PSY_3GPP_AH_THR_LONG    0.5f
 #define PSY_3GPP_AH_THR_SHORT   0.63f
 
+#define PSY_PE_FORGET_SLOPE  511
+
 enum {
     PSY_3GPP_AH_NONE,
     PSY_3GPP_AH_INACTIVE,
@@ -495,7 +497,7 @@ static int calc_bit_demand(AacPsyContext *ctx, float pe, int bits, int size,
     const float bitspend_add   = short_window ? PSY_3GPP_SPEND_ADD_S   : PSY_3GPP_SPEND_ADD_L;
     const float clip_low       = short_window ? PSY_3GPP_CLIP_LO_S     : PSY_3GPP_CLIP_LO_L;
     const float clip_high      = short_window ? PSY_3GPP_CLIP_HI_S     : PSY_3GPP_CLIP_HI_L;
-    float clipped_pe, bit_save, bit_spend, bit_factor, fill_level;
+    float clipped_pe, bit_save, bit_spend, bit_factor, fill_level, forgetful_min_pe;
 
     ctx->fill_level += ctx->frame_bits - bits;
     ctx->fill_level  = av_clip(ctx->fill_level, 0, size);
@@ -512,9 +514,14 @@ static int calc_bit_demand(AacPsyContext *ctx, float pe, int bits, int size,
      * Hopefully below is correct.
      */
     bit_factor = 1.0f - bit_save + ((bit_spend - bit_save) / (ctx->pe.max - ctx->pe.min)) * (clipped_pe - ctx->pe.min);
-    /* NOTE: The reference encoder attempts to center pe max/min around the current pe. */
+    /* NOTE: The reference encoder attempts to center pe max/min around the current pe.
+     * Here we do that by slowly forgetting pe.min when pe stays in a range that makes
+     * it unlikely (ie: above the mean)
+     */
     ctx->pe.max = FFMAX(pe, ctx->pe.max);
-    ctx->pe.min = FFMIN(pe, ctx->pe.min);
+    forgetful_min_pe = ((ctx->pe.min * PSY_PE_FORGET_SLOPE)
+        + FFMAX(ctx->pe.min, pe * (pe / ctx->pe.max))) / (PSY_PE_FORGET_SLOPE + 1);
+    ctx->pe.min = FFMIN(pe, forgetful_min_pe);
 
     /* NOTE: allocate a minimum of 1/8th average frame bits, to avoid
      *   reservoir starvation from producing zero-bit frames
diff --git a/tests/fate/aac.mak b/tests/fate/aac.mak
index bc9541e..2d41888 100644
--- a/tests/fate/aac.mak
+++ b/tests/fate/aac.mak
@@ -146,7 +146,7 @@ fate-aac-aref-encode: CMD = enc_dec_pcm adts wav s16le $(REF) -strict -2 -c:a aa
 fate-aac-aref-encode: CMP = stddev
 fate-aac-aref-encode: REF = ./tests/data/asynth-44100-2.wav
 fate-aac-aref-encode: CMP_SHIFT = -4096
-fate-aac-aref-encode: CMP_TARGET = 1127
+fate-aac-aref-encode: CMP_TARGET = 1139
 fate-aac-aref-encode: SIZE_TOLERANCE = 2464
 fate-aac-aref-encode: FUZZ = 6
 
@@ -191,7 +191,7 @@ fate-aac-is-encode: CMD = enc_dec_pcm adts wav s16le $(TARGET_SAMPLES)/audio-ref
 fate-aac-is-encode: CMP = stddev
 fate-aac-is-encode: REF = $(SAMPLES)/audio-reference/luckynight_2ch_44kHz_s16.wav
 fate-aac-is-encode: CMP_SHIFT = -4096
-fate-aac-is-encode: CMP_TARGET = 582
+fate-aac-is-encode: CMP_TARGET = 584
 fate-aac-is-encode: SIZE_TOLERANCE = 3560
 fate-aac-is-encode: FUZZ = 1