hashirama/uvg266
1
0
Fork 0
mirror of https://github.com/ultravideo/uvg266.git synced 2024-11-30 20:54:07 +00:00
Code Issues Actions Projects Releases Wiki Activity
4554 commits 3 branches 6 tags 10 MiB
Commit graph

302 commits

Author SHA1 Message Date
Joose Sainio ff02a84a96 Probably better order of things 2022-03-18 08:28:30 +02:00
Joose Sainio a9255901d9 Only perform rd2 calculation on the best candidate of the list if it exists 
But only for 2Nx2N blocks
 2022-03-18 08:28:29 +02:00
Joose Sainio a0e7165df4 use correct pu for rd calc 2022-03-18 08:28:29 +02:00
Joose Sainio 6d73db5a2a Probably correct RD cost calculation for all inter modes 2022-03-18 08:28:29 +02:00
Joose Sainio 1a9e54601f Calculate rd2 cost for all inter modes instead of just the final one 2022-03-18 08:28:29 +02:00
Joose Sainio f3f0037123 include root_cbf cost 2022-03-18 08:28:29 +02:00
Joose Sainio aaac260438 better merge cost 2022-03-18 08:28:28 +02:00
Joose Sainio 159793f5b4 more accurate get_mvd_coding_cost 2022-03-18 08:28:28 +02:00
Joose Sainio 5ed1ffb5d4 WIP: pre Christmas 2022-03-18 08:28:28 +02:00
Joose Sainio aea1133e6a Function for mock coding a CU and counting the bits 2022-03-18 08:27:22 +02:00
Joose Sainio d8648fe1de Preparation for accurately counting inter bits 2022-03-18 08:19:57 +02:00
Joose Sainio 311fceade7 Force use inter 2022-03-18 08:19:56 +02:00
Joose Sainio d5e4e831f4 Preliminary code for outputting bit costs during the search 2022-03-18 08:17:41 +02:00
Marko Viitanen 2545081be5 [threading] Added more margin to fracmv_within_tile() because of nondeterministic behaviour 2021-12-14 11:16:43 +02:00
Ari Lemmetti 51dd942778 Fix uninitialized fields of CU/PU infos. 2021-12-10 00:32:08 +02:00
Ari Lemmetti 2424a976a4 Prevent using uninitialized memory 2021-12-10 00:21:58 +02:00
Ari Lemmetti e45c6a9c68 Fix too few added keys in inter search stats. 
The function search_pu_inter_bipred may add more PUs than there are reference pictures.
 2021-12-10 00:03:37 +02:00
Ari Lemmetti a1a7036445 Fix indexing. Get rid of warning about jump depending on uninitialized value. 2021-12-08 23:06:39 +02:00
Ari Lemmetti bb1f2a0895 Reorder condition to prevent indexing past the array 2021-12-08 21:13:25 +02:00
Ari Lemmetti e87b12dec1 Move mv_cand initialization to better place 2021-12-08 21:09:16 +02:00
Ari Lemmetti f17a500b77 Get rid of warnings. (Unused variables, suggested parentheses) 2021-12-08 21:06:12 +02:00
Ari Lemmetti c411e65977 Prevent FME and bipred from the same reference picture if present in L0 and L1 2021-12-08 20:35:03 +02:00
Ari Lemmetti 49935710a8 Only one ME per reference picture (same ref in L0 and L1) 2021-12-08 16:35:47 +02:00
Ari Lemmetti ae498553c0 Add define MAX_UNIT_STATS_MAP_SIZE. Add assertions to inter search. 2021-12-08 15:53:31 +02:00
Ari Lemmetti 1af90b194e Add missing bits to RD costs. 2021-12-08 15:34:05 +02:00
Ari Lemmetti 706d718d5d Perform FME for n best PUs from L0 and L1. 2021-12-08 15:34:05 +02:00
Ari Lemmetti 7f7112cc57 Use up-to-date value of mv dir for bit cost calculations 2021-12-08 15:34:04 +02:00
Ari Lemmetti 4e19f7b71e Move mode decision logic and current PU setting to higher-level functions 2021-12-08 15:34:04 +02:00
Ari Lemmetti 3e967c0077 Add missing assertion and set cu before 2021-12-08 15:34:04 +02:00
Ari Lemmetti bdece66dc4 Compare the final costs only once and then set the current CU 2021-12-08 15:34:04 +02:00
Ari Lemmetti 0b223b24f2 Fix comment 2021-12-08 15:34:03 +02:00
Ari Lemmetti 4d02b69c4e Set CU type in inter search functions 2021-12-08 15:34:03 +02:00
Ari Lemmetti 2b9b398524 Remove now unnecessary state store/restore 2021-12-08 15:34:03 +02:00
Ari Lemmetti 3a219146ed Rename some variables 2021-12-08 15:34:03 +02:00
Ari Lemmetti dc4676eef1 Remove merge attempts from bipred functions 2021-12-08 15:34:03 +02:00
Ari Lemmetti adb31ce959 Use the new struct for bipred refinement as well 2021-12-08 15:34:02 +02:00
Ari Lemmetti 70a393a3dc Set mv candidates before cost calculations for bipred. Use the new struct for bipred. 2021-12-08 15:34:02 +02:00
Ari Lemmetti 574d6c4593 Eliminate copy-paste logic from different ME algorithms. 2021-12-08 15:34:02 +02:00
Ari Lemmetti d28c2295dc The best_* fields are no longer used to track anything. Convert costs to double. 2021-12-08 15:34:02 +02:00
Ari Lemmetti 9905cd42d6 Rename "indx" to "keys". There are too many "indices" already. 2021-12-08 15:34:01 +02:00
Ari Lemmetti 5edb82648a More intuitive logic for computing RD costs and bit costs for SMP 2021-12-08 15:34:01 +02:00
Ari Lemmetti aca9192054 Move cu_info_t initializations to search_pu_inter. Rename cur_cu cur_pu. 2021-12-08 15:34:01 +02:00
Ari Lemmetti 8406942d06 Improve the new data structure a bit. Use also for merge candidates. 2021-12-08 15:34:01 +02:00
Ari Lemmetti 94096dd175 Ignore merge candidates when computing AMVP motion vector costs. 2021-12-08 15:34:00 +02:00
Ari Lemmetti 48773b0d25 Replace and relocate deprecated cost and mode parameter tracking. 2021-12-08 15:34:00 +02:00
Ari Lemmetti 1940f0880f Add amvp unipredictions to both lists if reference picture is present 2021-12-08 15:33:57 +02:00
Ari Lemmetti 2ed434e57b Remove now deprecated array 2021-12-02 20:00:19 +02:00
Ari Lemmetti 90c0a708a7 Add new structs for storing statistics during the search. Use in AMVP search. 2021-12-02 20:00:19 +02:00
Ari Lemmetti 936fb76685 Remove merge candidate stuff from search_pu_inter_ref 
There is a separate merge analysis now
 2021-12-02 20:00:19 +02:00
Ari Lemmetti 3265d45a4e Temporarily remove FME threshold for verification purposes 2021-12-02 20:00:18 +02:00
Ari Lemmetti e000c7229f Fix bit costs in search_pu_inter_ref a bit 2021-12-02 20:00:18 +02:00
Ari Lemmetti ec2f4e0bac Use double for RD costs in most places 2021-12-02 20:00:15 +02:00
Marko Viitanen 3fda2dd6c8 [cleanup] Fix some warnings 2021-11-29 16:31:09 +02:00
Ari Lemmetti d6b2ec5814 Only check used reference picture lists when validating merge candidates. 
Merge candidate struct should be initialized to zero, so this should not have any effect. The conditions are added in case someone decides to copy the code as an example.
 2021-11-26 18:47:14 +02:00
Joose Sainio 385e91399a [intra rdo] Raise maximum rd level to 4 
Separate the rd option for chroma intra mode search from the full intra
mode search, i.e., rd=3 enables chroma mode search and rd=4 enables full
intra mode search.
 2021-11-26 08:12:37 +02:00
Marko Viitanen 348f54f5bb [tmvp] Change maximum MV in scaled mv clipping 2021-11-24 15:18:11 +02:00
Marko Viitanen 12e38684ff [inter] Change mvd coding cost to use cabac contexts 2021-11-23 16:25:31 +02:00
Marko Viitanen b644e85c1a [bipred] Fix bipred candidate selection 2021-11-23 10:58:47 +02:00
Marko Viitanen 54eaba3dee Fix merge issues and license texts 2021-11-23 08:46:06 +02:00
Marko Viitanen 885570e181 Merge branch 'master' of https://github.com/ultravideo/kvazaar 2021-11-22 22:11:32 +02:00
Marko Viitanen 1b86533b00 [inter] Scale MV's to qpel for cost calculation 2021-11-22 21:50:40 +02:00
Marko Viitanen 64705a27ee [fme] Enable 1/16th luma pixel fme 2021-11-22 20:31:30 +02:00
Marko Viitanen 33fc02d879 [me] Fix MV scaling in TZ search 2021-11-22 13:35:52 +02:00
Marko Viitanen 867304970e [fme] Enable avx2 interpolation and fix some warning about shifting mv's 
* Also switch mv int16_t to mv_t in many places
 2021-11-22 10:38:18 +02:00
Marko Viitanen d183c78fad [fme] Change fracmv_within_tile() to use internal MV resolution 2021-11-22 10:06:38 +02:00
Marko Viitanen ad8bad3f94 [inter] Scale x and y correctly in fracmv_within_tile() 2021-11-19 17:51:46 +02:00
Marko Viitanen 8841ed9c21 [inter] Fix fracmv_within_tile() assert to use correct mv resolution 2021-11-19 17:37:11 +02:00
Marko Viitanen 4d20461410 [inter] Limit merge search of some blocks with sub 1/4 pixel mv's 2021-11-19 17:28:13 +02:00
Marko Viitanen fa9a1db498 [inter] Fix mv precision in many places and add more mv_t usage and new vector2d_t rounding function 2021-11-19 16:20:49 +02:00
Marko Viitanen 5020f5f742 [inter] Fix incorrect mv scaling in unipred and change more mv types to mv_t 2021-11-18 11:49:08 +02:00
Marko Viitanen 21d7d2e4ed [inter] Remove MV rounding from kvz_inter_get_merge_cand() and add it to where merge cand are used 
* Should be adapted to AMVR later
 * merge candidates match VTM at full precision
 2021-11-18 11:09:26 +02:00
Marko Viitanen bf06538f33 [inter] Change internal MV precision to "INTERNAL_MV_PREC" and add new type mv_t 2021-11-18 09:49:12 +02:00
Marko Viitanen 30d97d9af6 [inter] Implement pairwise-average candidates for merge candidates 
- Half-pel candidates are skipped for now because it needs some special handling
 2021-11-01 13:24:23 +02:00
Marko Viitanen 57883369ca Change all the license texts in source headers and LICENSE file to 3-clause BSD, closes #302 
* All now have the same exact text string
 2021-10-13 15:22:46 +03:00
Marko Viitanen 7a5eb7712b Fix merge candidate derivation order 2021-10-08 16:34:02 +03:00
Marko Viitanen 99a2b0384d [cleanup] remove some warnings 2021-07-26 11:42:19 +03:00
Joose Sainio 1aaa95601c Merge remote-tracking branch 'remotes/kvz_github/master' into Fix-monochrome 
# Conflicts:
#	.gitlab-ci.yml
#	build/kvazaar_lib/kvazaar_lib.vcxproj.filters
#	src/cfg.c
#	src/encoder.h
#	src/kvazaar.h
#	src/rdo.c
 2021-04-23 10:56:50 +03:00
Ari Lemmetti b72ab583b4 Handle "don't care" rows in the end separately 2021-03-08 22:36:09 +02:00
Ari Lemmetti 475f1d79d5 Add some defines for important interpolation related sizes 2021-03-08 22:36:09 +02:00
Ari Lemmetti d8e7aac380 Do not use nonstandard extension for struct initialization. 2021-03-08 22:36:07 +02:00
Ari Lemmetti 84222cf3e7 Replace old block extrapolation with more capable one. 
Separate paddings for different directions can be now specified.
 2021-03-08 22:36:04 +02:00
Arttu Makinen ebb99a7223 Changed 'width's to 'stride's, because added more pixels to 'fulldata'. 
Also some small fixes and changes.
Checksum correct in luma.
WIP
 2020-12-30 14:30:47 +02:00
Ari Lemmetti d1abf85229 Add MV constraint check to motion estimation start point 2020-06-01 23:51:38 +03:00
Ari Lemmetti f31dddc019 Bypass inverse quantization and inverse transform when trying early skip 2020-04-10 16:02:09 +03:00
RLamm 27fe716654 Fixed reference POC indexing 2020-03-11 15:33:37 +02:00
RLamm bf24831780 Attempt to fix random crashes 2020-03-11 15:31:47 +02:00
RLamm 887659db1f Attempted to scale the extra_mvs 2020-03-11 15:31:46 +02:00
Ari Lemmetti 1dfc69b42e Consider merge index bits in merge analysis and early skip 2020-02-20 09:43:58 +02:00
Ari Lemmetti 71d9327f62 Further improve fast bipred 2020-02-17 20:32:52 +02:00
Ari Lemmetti 886ff36d12 Initial implementation of fast bipred. 2020-02-04 15:46:23 +02:00
Ari Lemmetti 147378e1f9 Prevent 8x4 and 4x8 bipred in merge analysis 2019-09-03 16:32:50 +03:00
Ari Lemmetti ef1fdbf259 Separate prediction of single PU/PB from CU/CB 2019-09-03 16:32:50 +03:00
Ari Lemmetti 3bc510712f Enable merge analysis for smp and amp 2019-09-02 17:31:51 +03:00
Ari Lemmetti 557bcbc6aa Make luma or chroma only inter "recon" or predict possible 2019-09-02 17:15:28 +03:00
Ari Lemmetti aa8ab195d1 Compare rough cost of the best merge mode against AMVP to make mode decision 2019-08-26 22:49:09 +03:00
Ari Lemmetti 8f866ff83a Use correct index 2019-08-26 20:10:10 +03:00
Ari Lemmetti a80de22bc7 Add only different candidates to the list 2019-08-24 21:49:33 +03:00
Ari Lemmetti 45c7961412 Remove tr depth fill. It should not be needed. 2019-08-24 21:49:32 +03:00
Ari Lemmetti ff8711aaab Add missing logic to add valid indices to list 2019-08-24 21:49:29 +03:00
Ari Lemmetti b20992a9f3 Rename functions more descriptive 2019-07-09 20:20:11 +03:00