Gregory Maxwell authored 13 years ago and Jean-Marc Valin committed 13 years ago

non-ascii characters from the source.

Got authorization from all copyright holders

This was computing on uninitialized data (though the result was never
used)

This provides more entropy and allows some more flexibility on the
encoder side.

Timothy B. Terriberry authored 14 years ago and Jean-Marc Valin committed 14 years ago

Jean-Marc's original anti-collapse patch used a threshold on the
 content of a decoded band to determine whether or not it should
 be filled with random noise.
Since this is highly sensitive to the accuracy of the
 implementation, it could lead to significant decoder output
 differences even if decoding error up to that point was relatively
 small.

This patch detects collapsed bands from the output of the vector
 quantizer, using exact integer arithmetic.
It makes two simplifying assumptions:
 a) If either input to haar1() is non-zero during TF resolution
     adjustments, then the output will be non-zero.
 b) If the content of a block is non-zero in any of the bands that
     are used for folding, then the folded output will be non-zero.
b) in particular is likely to be false when SPREAD_NONE is used.
It also ignores the case where mid and side are orthogonal in
 stereo_merge, but this is relatively unlikely.
This misses just over 3% of the cases that Jean-Marc's anti-collapse
 detection strategy would catch, but does not mis-classify any (all
 detected collapses are true collapses).

This patch overloads the "fill" parameter to mark which blocks have
 non-zero content for folding.
As a consequence, if a set of blocks on one side of a split has
 collapsed, _no_ folding is done: the result would be zero anyway,
 except for short blocks with SPREAD_AGGRESSIVE that are split down
 to a single block, but a) that means a lot of bits were available
 so a collapse is unlikely and b) anti-collapse can fill the block
 anyway, if it's used.
This also means that if itheta==0 or itheta==16384, we no longer
 fold at all on that side (even with long blocks), since we'd be
 multiplying the result by zero anyway.

There's no explicit signaling for it yet.

Folding now moved to alg_quant() and alg_unquant(). No more funny short block
alignment checks because bands now have to be aligned.

now that we no longer do normalized pitch

tonal artefacts. This replaces folding for all cases where we have at least
one pulse.

for M and S.

commit ea807b68678dd76175def2c5eb006c6bdb16679e
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Wed Mar 25 23:24:41 2009 -0400

    cleanup before merge

commit 73ad1a0202641be6a23903e464ece21fe332a131
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sat Mar 21 00:02:16 2009 -0400

    Some tuning of the new stereo

commit c05057eb57a7723045214a2f830fd561388ae48a
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sun Mar 15 19:56:11 2009 -0400

    fixed-point: atan2() converted

commit a8476cf8be55b0612d42df98d9807ca1335adfe3
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sat Mar 14 23:10:04 2009 -0400

    fixed-point: Getting the new stereo code working in fixed-point (still more
    work left)

commit 70a452761a5ce15700664e7167886dce5914cbd0
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Mon Mar 2 23:36:25 2009 -0500

    Coding left and right independently for lower bands

commit 4efd1e6385c7d036749080265a8d26668312b91b
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sun Mar 1 23:56:46 2009 -0500

    Removed the sqrt(C) from the normalisation, which simplifies a lot of things.

commit a4f3c5c60bc396bf644afa49b49e6b24ccf144f8
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Fri Feb 20 20:49:38 2009 -0500

    Better point stereo calculation when we don't encode the side anyway

commit f08525de4739f4017d19ec2e2022883deda8f826
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Thu Feb 19 19:43:09 2009 -0500

    Apparently, Timothy's calculations for fine energy allocation also apply
    to the quantisation of theta.

commit 6548cffc9d3f996b8a8dbfab982f0da0bc6c2dc2
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Thu Feb 19 07:35:24 2009 -0500

    Better handling of the "theta bits" and disabling the orthogonalize()

commit 7aa82c694967afa85dd30be8cc670663f98829fe
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Wed Feb 18 08:01:07 2009 -0500

    tuning the new stereo

commit c2b780a773de66fd9613c7cd54c09b705fe9ce45
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sun Feb 15 22:24:52 2009 -0500

    The new stereo coupling actually decodes properly now.

commit 85513c203d773bebcf0a6055f953170d563d890c
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date:   Sun Feb 15 21:31:16 2009 -0500

    First attempt at a new "constrained" MS stereo scheme

process

an offset.

can be optimised

a division

the more complex "n-best search" and replaced it with a greedy search

converted to float though)

copyright, so it's fine!)