Commit Graph

1727 Commits

Author SHA1 Message Date
skal
5935259cd0 add comment about the kLevelsFromDelta[][] LUT generation
Change-Id: Id1a91932bf05aeee421761667af351ef2200c334
2013-10-31 21:35:55 -07:00
skal
e3312ea681 detect flatness in blocks and favor DC prediction
this avoids local-minima that look bad, even if the distortion
looks low (e.g. gradients, sky,...). Mostly visible in the q=50-80 range.
Output size is mostly unchanged.

Change-Id: I425b600ec45420db409911367cda375870bc2c63
2013-11-01 00:47:04 +01:00
James Zern
ebc9b1eedf Merge "VPLBitReader bugfix: Catch error if bit_pos > LBITS too." 2013-10-31 16:38:15 -07:00
Urvang Joshi
96ad0e0aef VPLBitReader bugfix: Catch error if bit_pos > LBITS too.
Earlier we were only testing for bit_pos == LBITS. But this is not
sufficient,
as bit_pos can jump from < LBITS to > LBITS.

This was resulting in some bit-stream truncation errors not being
caught.

Note: Not a security bug though, as br->pos wasn't incremented in such
cases
and so we weren't reading beyond the buffer.

Change-Id: Idadcdcbc6a5713f8fac3470f907fa37a63074836
2013-10-30 16:33:36 -07:00
skal
a014e9c9cd tune quantization biases toward higher precision
* raise U/V quantization bias to more neutral values
* also raise the non-zero AC bias for Y1/Y2 matrices

(we need all the precision we can for U/V leves, which are often empty)
This will increase quality in the higher range (q >= 90) mostly.
Files size is exacted to raise a little (5-7%). and SSIM accordingly of course.

Change-Id: I8a9ffdb6d8fb6dadb959e3fd392e66dc5aaed64e
2013-10-30 23:57:23 +01:00
skal
1e898619cb add helpful PrintBlockInfo() function
(protected under a DEBUG_BLOCK compile flag)

Change-Id: Icb8da02dbd00e5cf856c314943c212f1c9578d9b
2013-10-30 19:25:27 +01:00
Pascal Massimino
596a6d73ce make use of 'extern' consistent in function declarations
Change-Id: I18e050db3111e52acfe97da09cdf1860f3e15936
2013-10-30 03:23:21 -07:00
Pascal Massimino
c8d48c6e0d Merge "extract random utils to their own file util/random.[ch]" 2013-10-30 02:07:28 -07:00
Pascal Massimino
98aa33cf1e extract random utils to their own file util/random.[ch]
they'll be used for decoding too, probably.

Change-Id: Id9cbb250c74fc0e876d4ea46b1b3dbf8356d6725
2013-10-30 02:00:33 -07:00
pascal massimino
432a723e0b Merge "swig: add basic go bindings" 2013-10-30 01:03:32 -07:00
pascal massimino
fab618b5fb Merge "rename libwebp.i -> libwebp.swig" 2013-10-30 01:02:05 -07:00
James Zern
e4e7fcd69b swig: add basic go bindings
WebPGetDecoderVersion
WebPGetInfo

Change-Id: Icb66207c2f492c9d20642544c0b31cc92b91cdcc
2013-10-29 21:44:20 +01:00
James Zern
d3408720d8 Merge "fast auto-determined filtering strength" 2013-10-29 12:47:58 -07:00
skal
f8bfd5cd1e fast auto-determined filtering strength
kLevelsFromDelta[sharpness][delta] is an inverse look-up table
that tells the minimum filtering strength needed to trigger the
filtering of a step with amplitude 'delta'. We use this table
in various situations:

a) when computing the initial (/global) filtering
strength for each segment. We look at the quantization
step and deduce the proper filtering strength needed
to result this quantization noise (talking the -f option
into account).

b) during intra16 calculation, when a block ends up
very empty (only DC coeffs are non-zero, all ACs have
vanished). We'll rely on the in-loop filtering to
restore the smoothness (if the source was gradient-like
smooth. That's why we look at the distortion too before
triggering the filtering).

Step b) goes _in addition_ to a), potentially raising
the filtering strength if blockiness is likely.

Change-Id: Icaeca93ef21da195b079e6587a44d9edfc8e9efa
2013-10-29 20:13:29 +01:00
Pascal Massimino
ac0bf951ca small clean-up in ExpandMatrix()
Change-Id: Ib06cb1658a6548f06bb7320310b3864881b606a7
2013-10-29 19:58:57 +01:00
James Zern
1939607e7f rename libwebp.i -> libwebp.swig
.swig provides better tool compatibility

Change-Id: I35587b84dd32e12e8cedad14b36ac518b63896bc
2013-10-29 19:51:04 +01:00
skal
43148b6cd2 filtering: precompute ilimit and hev_threshold
no speed change, just simplifying the logic

Change-Id: I518800494428596733d4fbae69072049828aec3c
2013-10-28 13:37:33 +01:00
Pascal Massimino
18f992ec0f simplify f_inner calculation a little
by incorporating the is_4x4 flag at init

Change-Id: I042e04aacb15181db0bf86f3212c880087519189
2013-10-28 01:49:09 -07:00
Pascal Massimino
241d11f141 add missing const
Change-Id: Id1c767d21d52197ed2e4497005eb9c4795c602f0
2013-10-25 20:34:14 +02:00
Pascal Massimino
86c0031eb2 add a 'format' field to WebPBitstreamFeatures
Change-Id: I79a688e4c34fb77527127bbdf4bc844efa6aa9a4
2013-10-25 20:34:06 +02:00
Urvang Joshi
dde91fde96 Demux: Correct the extended format validation
Earlier "f = f->next_" was executing for both inner and outer loop, thus
skipping validation of some frames.

Change-Id: Ice5cdb4ff5da78384aa0573addd3a5e5efa0b10c
2013-10-23 17:25:23 -07:00
Pascal Massimino
5d6c5bd2a5 add entry for '-resize' option in cwebp's man
Change-Id: I04b2853825527b03cfa73377004b1a2a40b4c94f
2013-10-23 02:22:15 -07:00
skal
7c098d1814 Use some gamma-curve range compression when computing U/V average
This helps for discolorated chroma-subsampled edges.

Change-Id: I1d8ce87b66cb7e8b3572e6722905beabf0f50554
2013-10-18 21:28:05 +02:00
skal
0b2b05049f Use deterministic random-dithering during RGB->YUV conversion
-> helps debanding (sky, gradients, etc.)

This dithering can only be triggered when using -preset photo
or -pre 2 (as a preprocessing). Everything is unchanged otherwise.

Note that this change is likely to make the perceived PSNR/SSIM drop
since we're altering the input internally.

Change-Id: Id8d4326245d9b828141de162c94ba381b1fa5813
2013-10-17 22:36:49 +02:00
skal
8a2fa099cc Add a second multi-thread method
method 1 grouping: [parse + reconstruction] // [filtering + output]
method 2 grouping: [parse] // [reconstruction+filtering + output]

Depending on some heuristics (see VP8ThreadMethod()), we
can pick one of the other when -mt flag (or option.use_threads)
is selected.

Conservatively, we always use method #2 for now until the heuristic
is refined (so, timing should be the same the before this patch)

+ replace 'use_threads' by 'mt_method'
+ define MIN_WIDTH_FOR_THREADS constant
+ fix comment alignment

Change-Id: I11a756dea9070d6e21b1a9481d357a1e8aa0663e
2013-10-15 23:58:31 +02:00
skal
7d6f2da075 Merge "up to 20% faster multi-threaded decoding" 2013-10-15 14:51:12 -07:00
James Zern
266f63ea89 Merge "libwebp.jar: build w/Java 1.6 for Android compat" 2013-10-15 02:04:41 -07:00
skal
0532149c8a up to 20% faster multi-threaded decoding
Mostly visible for large images.
Reconstruction+filtering is now done in parallel to bitstream-parsing.

Change-Id: I4cc4483d803b255f4d97a2fcd9158b1c291dd900
2013-10-15 00:25:21 +02:00
Urvang Joshi
38efdc2e9c Simplify the gif2webp tool: move the optimization details to util
Specifically:
- Merge OptimizeAndEncodeFrame with WebPFrameCacheAddFrame: they use the same
  if-else structure.
- Move maintenance of 'prev_canvas' and 'curr_canvas' to util.
- Move ReduceTransparency() and FlattenPixels() calls to SetFrame(): This is in
  preparation for the next patch: which will try try lossless encoding for
  each frame, even when '-lossy' option is given.
- Make most methods static inside util.

No changes to output expected.

Change-Id: I1f65af25246665508cb20f0f6e338f9aaba9367b
2013-10-14 14:39:46 -07:00
James Zern
de899516c7 libwebp.jar: build w/Java 1.6 for Android compat
broken since:
 a5c297c swig/java: reduce wrapper function code duplication

this was a part of v0.3.1, but not v0.3.0.

Change-Id: I001d4bd0a7a1aa1b2d267bc63bc1d8226bff00c1
2013-10-11 17:59:37 +02:00
skal
cb22155201 Decode a full row of bitstream before reconstructing
Needs more memory but allows for future parallelization.
Noticeably faster on ARM, slightly faster on x86

also: remove dec->filter_row_ unnecessary field

Change-Id: I044a808839b4e000c838a477e3e8688820436d9a
2013-10-10 21:29:58 +02:00
James Zern
dca8a4d315 Merge "NEON/simple loopfilter: avoid q4-q7 registers" 2013-10-10 01:58:41 -07:00
pascal massimino
9e84d901d2 Merge "NEON/TransformWHT: avoid q4-q7 registers" 2013-10-09 09:32:59 -07:00
James Zern
fc10249b36 NEON/simple loopfilter: avoid q4-q7 registers
very tiny speed improvement

Change-Id: I3024f120feb7275ce20bfff21af31ea8650a5a03
2013-10-09 18:17:31 +02:00
James Zern
2f09d63e30 NEON/TransformWHT: avoid q4-q7 registers
very tiny speed improvement

Change-Id: Iace78b9038af412d0a794845ff19f54afa88ccdc
2013-10-09 18:17:23 +02:00
skal
77585a2be6 Merge "use a macrofunc for setting NzCoeffs bits" 2013-10-09 03:00:49 -07:00
skal
d155507cb9 Merge "use HINT_GRAPH as image_hint for gif source" 2013-10-09 02:58:15 -07:00
skal
9c561646ac Merge "only print GIF_DISPOSE_WARNING once" 2013-10-09 02:56:45 -07:00
skal
05879865da use HINT_GRAPH as image_hint for gif source
quite rarely, it gives a different output
then without the HINT, but that's often for
a smaller size (tested with default -m and -m 6)

Change-Id: I51d221ab61f8e007983325031345728e8d80b241
2013-10-09 11:55:51 +02:00
skal
0b28d7ab08 use a macrofunc for setting NzCoeffs bits
(avoids code dup)

Change-Id: I776f065538e562673ca08f3bc43c7167d13254d9
2013-10-09 11:46:32 +02:00
skal
f9bbc2a034 Special-case sparse transform
If the number of non-zero coeffs is <= 3, use a
simplified transform for luma.

Change-Id: I78a1252704228d21720d4bc1221252c84338d9c8
2013-10-08 22:05:38 +02:00
skal
00125196f3 gif2webp: detect and flatten uniformly similar blocks
helps during lossless compression.

10% average saving, but that's mostly on what was previously
'difficult' cases, where the gain is ~30-50% actually.
Non-difficult cases are mostly unchanged.
Tested over ~7k random web gifs.

Change-Id: I09db4560e4ab09105d1cad28e6dbf83842eda8e9
2013-10-08 15:04:52 +02:00
skal
0deaf0fa44 only print GIF_DISPOSE_WARNING once
Change-Id: Iece235a6ee767cc2e3866bb6c9bb1d56d9a9ffb9
2013-10-08 15:02:33 +02:00
Pascal Massimino
6a8c0eb718 Merge "small optimization in segment-smoothing loop" 2013-10-08 04:13:37 -07:00
Pascal Massimino
f7146bc1e6 small optimization in segment-smoothing loop
probably not much of a speed difference

Change-Id: I08c41d82c3c2eb5ff9ec9ca9d81af2bb09b362de
2013-10-07 07:44:51 -07:00
skal
5a7533ceb3 small gif2webp fix
the old kmin value was printed instead of the new one.

Change-Id: I0dcf23a29b7ce6bad08013a1980c5ce027b1bfe9
2013-10-04 21:13:35 +02:00
James Zern
4df0c89e24 Merge changes Ic697660c,I27285521
* changes:
  Android.mk: add a dwebp target
  Android.mk: update build flags
2013-10-04 05:02:13 -07:00
James Zern
5b2e6bd3e8 Android.mk: add a dwebp target
Change-Id: Ic697660c1a5f7185d2ad00934b314b44870cec00
2013-10-04 11:26:37 +02:00
James Zern
f910a84ea5 Android.mk: update build flags
- split out release specific flags
- set LOCAL_ARM_MODE to arm

Change-Id: I272855216583d6c8d0a4106e8b3fde46aa59dfa9
2013-10-04 11:18:09 +02:00
skal
63f9aba4b3 special-case WHT transform when there's only DC
happens surprisingly often at low quality, so we might
as well hard-code a simplified TransformWHT() directly.

Change-Id: Ib7a858ef74e8f334bd59d6512bf5bd3e455c5459
2013-10-02 14:27:36 +02:00