Re: [x265] Please try to visualize motion search methods
Am 05.12.2016, 05:53 Uhr, schrieb Pradeep Ramachandran: A recent post on doom9 points to this post that shows HEX/DIA/FULL search. http://forum.doom9.org/showthread.php?p=693742 Yes, there are a few examples. But I don't just ask for me alone. I ask for everyone who would be interested in understanding HEVC in general and x265 specifically. So I hope you will be able to display such images (and possibly even animations or color-layered images in cases of UMH and SEA?) on your web documentation (readthedocs). If I already understood them all, I would even volunteer to create them. Unfortunately, I do not yet. But I can imagine that it will help imagining the tradeoffs between speed and quality, in relation to the probability of motion directions and coverage of a neighborhood of a currently discovered pixel/macroblock/CU ... I am already unsure about the base unit. Well, certainly not pixel. -- Fun and success! Mario *LigH* Rohkrämer mailto:cont...@ligh.de ___ x265-devel mailing list x265-devel@videolan.org https://mailman.videolan.org/listinfo/x265-devel
[x265] [PATCH] regression: add pme option in regression commandlines
# HG changeset patch # User Mahesh Pittala# Date 1480916513 -19800 # Mon Dec 05 11:11:53 2016 +0530 # Node ID b2d360143d966de3832a05d8af1c824ecb5b8c5f # Parent f557d3ba58c2fa54629476137ea64535f9547b3a regression: add pme option in regression commandlines removed pme, pmode options from spot options(randomly added in regression commandlines) and added here. diff -r f557d3ba58c2 -r b2d360143d96 source/test/regression-tests.txt --- a/source/test/regression-tests.txt Fri Dec 02 11:20:51 2016 +0530 +++ b/source/test/regression-tests.txt Mon Dec 05 11:11:53 2016 +0530 @@ -26,6 +26,7 @@ BasketballDrive_1920x1080_50.y4m,--preset veryslow --recon-y4m-exec "ffplay -i pipe:0 -autoexit" Coastguard-4k.y4m,--preset ultrafast --recon-y4m-exec "ffplay -i pipe:0 -autoexit" Coastguard-4k.y4m,--preset superfast --tune grain --overscan=crop +Coastguard-4k.y4m,--preset superfast --tune grain --pme --aq-strength 2 --merange 190 Coastguard-4k.y4m,--preset veryfast --no-cutree --analysis-mode=save --bitrate 15000,--preset veryfast --no-cutree --analysis-mode=load --bitrate 15000 Coastguard-4k.y4m,--preset medium --rdoq-level 1 --tune ssim --no-signhide --me umh --slices 2 Coastguard-4k.y4m,--preset slow --tune psnr --cbqpoffs -1 --crqpoffs 1 --limit-refs 1 @@ -71,11 +72,13 @@ News-4k.y4m,--preset superfast --slices 4 --aq-mode 0 News-4k.y4m,--preset medium --tune ssim --no-sao --qg-size 16 News-4k.y4m,--preset veryslow --no-rskip +News-4k.y4m,--preset veryslow --pme --crf 40 OldTownCross_1920x1080_50_10bit_422.yuv,--preset superfast --weightp OldTownCross_1920x1080_50_10bit_422.yuv,--preset medium --no-weightp OldTownCross_1920x1080_50_10bit_422.yuv,--preset slower --tune fastdecode ParkScene_1920x1080_24_10bit_444.yuv,--preset superfast --weightp --lookahead-slices 4 ParkScene_1920x1080_24.y4m,--preset medium --qp 40 --rdpenalty 2 --tu-intra-depth 3 +ParkScene_1920x1080_24.y4m,--preset medium --pme --tskip-fast --tskip --min-keyint 48 --weightb --limit-refs 3 ParkScene_1920x1080_24.y4m,--preset slower --no-weightp RaceHorses_416x240_30.y4m,--preset superfast --no-cutree RaceHorses_416x240_30.y4m,--preset medium --tskip-fast --tskip ___ x265-devel mailing list x265-devel@videolan.org https://mailman.videolan.org/listinfo/x265-devel
Re: [x265] Please try to visualize motion search methods
On Mon, Dec 5, 2016 at 1:41 AM, Mario Rohkrämerwrote: > Only you, as the developers, will know which geometrical shapes the > different motion search methods will use, only you can give us a visual > impression how many samples in which direction and which distance will be > considered for each method. > > I tried for a long time to find diagrams in the web, but this is a topic > where my experience with Google image search fails. "A picture is worth a > thousand words" - this proverb fits here better than ever. > A recent post on doom9 points to this post that shows HEX/DIA/FULL search. http://forum.doom9.org/showthread.php?p=693742 > > > And, by the way, I believe the CLI help is improvable for this line: > >--me Motion search method dia hex umh star > full. Default 1 > > a) There should be a delimiter between the explanation and the set of > valid options; I guess you will have a common syntax. > > b) The default value is an integer here, but the option type is documented > as "string". > > -- > > Fun and success! > Mario *LigH* Rohkrämer > mailto:cont...@ligh.de > ___ > x265-devel mailing list > x265-devel@videolan.org > https://mailman.videolan.org/listinfo/x265-devel > ___ x265-devel mailing list x265-devel@videolan.org https://mailman.videolan.org/listinfo/x265-devel
[x265] Please try to visualize motion search methods
Only you, as the developers, will know which geometrical shapes the different motion search methods will use, only you can give us a visual impression how many samples in which direction and which distance will be considered for each method. I tried for a long time to find diagrams in the web, but this is a topic where my experience with Google image search fails. "A picture is worth a thousand words" - this proverb fits here better than ever. And, by the way, I believe the CLI help is improvable for this line: --me Motion search method dia hex umh star full. Default 1 a) There should be a delimiter between the explanation and the set of valid options; I guess you will have a common syntax. b) The default value is an integer here, but the option type is documented as "string". -- Fun and success! Mario *LigH* Rohkrämer mailto:cont...@ligh.de ___ x265-devel mailing list x265-devel@videolan.org https://mailman.videolan.org/listinfo/x265-devel