diff options
| author | Rostislav Pehlivanov <atomnuker@gmail.com> | 2015-09-12 13:04:27 +0100 |
|---|---|---|
| committer | Rostislav Pehlivanov <atomnuker@gmail.com> | 2015-09-12 13:04:27 +0100 |
| commit | a83a8d70681a7093ff4bd8232b8f75fed3d6af48 (patch) | |
| tree | a28f3bc5df2467018713f86169a622150fdce7a9 /libavcodec/aacenc_tns.h | |
| parent | 33be39efe31326dc787efc54f5077e4c4b97896c (diff) | |
aacenc_tns: redo coefficient quantization and decision making
This finally (and again) gets rid of basically everything the
specifications say about how TNS should be done. The main
problem used to be that a single filter was used for all
coefficients which despite being explicitly recommended by
the specifications usually sounds wrong, therefore it's
a corner case in the current TNS implementation.
This commit also changes the coefficient bit size, as apparently
it's better to use lower precision in case the windows are eight
short. This is apparently what fdk_aac uses, looking at the bit
stream and makes sense. Also the order when 8 SHORT windows happen
is important as 7 was too much and according to PSNR was worse
while 5 is just about correct.
Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Diffstat (limited to 'libavcodec/aacenc_tns.h')
| -rw-r--r-- | libavcodec/aacenc_tns.h | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/libavcodec/aacenc_tns.h b/libavcodec/aacenc_tns.h index f0e75569ae..46e4794965 100644 --- a/libavcodec/aacenc_tns.h +++ b/libavcodec/aacenc_tns.h @@ -33,6 +33,9 @@ /* Could be set to 3 to save an additional bit at the cost of little quality */ #define TNS_Q_BITS 4 +/* Coefficient resolution in short windows */ +#define TNS_Q_BITS_SHORT 3 + /* TNS will only be used if the LPC gain is within these margins */ #define TNS_GAIN_THRESHOLD_LOW 1.395f #define TNS_GAIN_THRESHOLD_HIGH 11.19f |