Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
Why 10 bit YUV compresses better than 8 bit YUV in x264 is complex and I don't understand it.
It is easiest for me to think of it in terms of banding.
A color gradient in 8 bit has banding because we can see the difference from one color to the next.
To fix this in 8 bit you can apply dithering, a checkerboard pattern which gives an appearance of a greater number of colors.
But the pattern is hard to compress, so you lose it if you don't up the bitrate.
10 bit has more colors, so it doesn't need dithering, so you can produce the same visual effect in less bits.
(you don't have a 10bit monitor so the dithering is still there, but it is added after decompression by the codec or your graphics card)
Chroma Subsampling
Dropping most of the color data cuts the raw file size/bandwidth in half.
This works ok because color data is mostly redundant.
But video compression is great at reducing exactly the type of redundancy present in chroma data.
So for compressed video you can have full chroma at very little increase to bitrate.
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
creaothceann wrote:
What I noticed is that Deblink doesn't show shaking in the exploding-Mother-Brain scene of the intro, so there's still no single algorithm for everything.
Yah, frustration over that scene is why I gave up working on this stuff the last time.
It is in black/white (so color is no help for matching), the intensity changed with time, and the shake pattern skips a frame in the middle =_=
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
If {1 frame off, 1 frame on} at 60 fps looks like transparency to you, your monitor sucks.
Yes, that is definitely a possibility =_=
creaothceann wrote:
nanogyth wrote:
Still plenty of artifacts, but I think it is a step in the right direction
Do you think they're too obvious or that you can fix them? If not then it could be added to the site as an alternative to 50/50 blending and TASBlend.
It is an imperfect solution to a hard problem. A useful tool that can complement the others, but not something which should be blindly applied in all situations.
First thing to add is functionality similar to TASblend
function deblink2(clip clp, float "ratio"){
ratio = default(ratio, 2.0 / 3)
opacity1 = round((1 - ratio) * 257)
opacity2 = round(( ratio) * 257)
# blend on the frame it blinks in and out
blink=clp.ng_blinkmask_new
blink2=mt_logic(blink.selectevery(1,-1), blink, mode="or").converttorgb32
even = layer(clp, clp.trim(1,0).mask(blink2), level=opacity1)
odd = layer(clp, clp.trim(1,0).mask(blink2), level=opacity2)
interleave(even.selectevery(4,0),even.selectevery(4,1),
\ odd.selectevery(4,2),odd.selectevery(4,3))
}
That seems to work pretty well. Can slide between no flicker at .5 and full flicker at 1.0
Then there are some issues with the blinkmask not picking up all the motion. (roadrunners toes get chopped off)
function deblink3(clip clp, float "ratio", int "radius"){
radius = default(radius, 1)
ratio = default(ratio, 2.0 / 3)
opacity1 = round((1 - ratio) * 257)
opacity2 = round(( ratio) * 257)
# blend on the frame it blinks in and out
blink=clp.ng_blinkmask_new
blink2=mt_logic(blink.mt_expand(mode=mt_circle(radius)), clp.shake, mode="and")
blink3=mt_logic(blink, blink2, mode="or")
blink4=mt_logic(blink3.selectevery(1,-1), blink3, mode="or")
masked_clp=clp.trim(1,0).mask(blink4.converttorgb32)
even = layer(clp, masked_clp, level=opacity1)
odd = layer(clp, masked_clp, level=opacity2)
interleave(even.selectevery(4,0),even.selectevery(4,1),
\ odd.selectevery(4,2),odd.selectevery(4,3))
}
function shake(clip c){
e0=c.selectevery(1,0)
o1=c.selectevery(1,1)
e2=c.selectevery(1,2)
mt_logic(ng_diff(e0,o1),ng_same(e0,e2),mode="and")
}
function ng_diff(clip A, clip B, int "thr"){
thr=default(thr,0)
TAD=ng_TAD(A,B)
return mt_binarize(TAD, threshold=thr)
}
function ng_same(clip A, clip B, int "thr"){
thr=default(thr,0)
TAD=ng_TAD(A,B)
return mt_binarize(TAD, threshold=thr, upper=true)
}
function ng_TAD(clip A, clip B){
R=ng_AD(A .showRed("YV12"),B .showRed("YV12"))
G=ng_AD(A.showGreen("YV12"),B.showGreen("YV12"))
B=ng_AD(A .showBlue("YV12"),B .showBlue("YV12"))
return ng_plus(R, ng_plus(G, B))
}
function ng_AD(clip A, clip B){
return mt_lutxy(A,B,"x y - abs")
}
function ng_plus(clip A, clip B){
return mt_lutxy(A,B,"x y +")
}
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
natt wrote:
but I think the expected user viewing result with a lot of these flicker patterns was in fact flicker.
To me, at 60fps, "on one, off one" looks like transparency and "on two, off two" looks like flickering.
Changing both to "on one, off one" at 30fps is the wrong thing to aim for IMO.
If you download the roadrunner clip it has the three side by side (on/off, 60fps, blended). I think the blended looks more like the 60fps, would you disagree?
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
Use of flickering in games is usually used to simulate transparency, not flickering.
That works at 60fps, but at 30fps it isn't right.
My idea was to blend one frame effects into two frame effects.
Making a 60fps clip which is safe to reduce to 30fps.
Still plenty of artifacts, but I think it is a step in the right direction, do you agree?
http://www.youtube.com/watch?v=tjrRqrOQ1O8http://www.mediafire.com/?3l6x53xq9999q8m
function deblink(clip clp){
# blend on the frame it blinks in and out
blink=clp.ng_blinkmask_new
blink2=mt_logic(blink.selectevery(1,-1), blink, mode="or").converttorgb32
layer(clp, clp.trim(1,0).mask(blink2), level=127)
}
function ng_blinkmask_new(clip c,int "ml"){
ml=default(ml,128)
src=c.ConvertToYv12
super=MSuper(src, pel=1)
fvec =MAnalyse(super, isb=false, blksize=4)
bvec =MAnalyse(super, isb=true , blksize=4)
fmask=Mmask(src,fvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
bmask=Mmask(src,bvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
eo0_to =fmask.selectevery(2,1)
oe_from=bmask.selectevery(2,1)
front =mt_logic(eo0_to,oe_from,mode="and")
oe_to =fmask.selectevery(2,2)
eo_from=bmask.selectevery(2,2)
back =mt_logic(oe_to,eo_from,mode="and")
ee_src=src.selecteven
ee_super=MSuper(ee_src, pel=1)
ee_fvec =MAnalyse(ee_super, isb=false, blksize=4)
ee_bvec =MAnalyse(ee_super, isb=true , blksize=4)
ee_fmask=Mmask(ee_src,ee_fvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
ee_bmask=Mmask(ee_src,ee_bvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
ee_to =ee_fmask.trim(1,0)
ee_from=ee_bmask
ee =mt_logic(ee_to,ee_from,mode="or")
oo_src=src.selectodd
oo_super=MSuper(oo_src, pel=1)
oo_fvec =MAnalyse(oo_super, isb=false, blksize=4)
oo_bvec =MAnalyse(oo_super, isb=true , blksize=4)
oo_fmask=Mmask(oo_src,oo_fvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
oo_bmask=Mmask(oo_src,oo_bvec,kind=1,ml=ml).mt_binarize(u=-128,v=-128)
oo_to =oo_fmask.trim(1,0)
oo_from=oo_bmask
oo =mt_logic(oo_to,oo_from,mode="or")
#to e0-o1, from o1-e2, nothing e0-e2
even_blink=mt_logic(front,ee.mt_invert,mode="and")
#to o1-e2, from e2-o3, nothing o1-o3
odd_blink =mt_logic(back,oo.mt_invert,mode="and")
blinkmask=interleave(even_blink, odd_blink)
return blinkmask
}
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
Why not have lossless RGB as the primary encode?
The quality would be better, and it would fail properly if you didn't have it setup right.
I've heard lossless is smaller than x264 compressed on some games, is that generally true?
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
These are some tools I put together when I was hacking around at encoding final fantasy viii. I'm sharing in case someone else might find them useful, or someone can give me some tips on my usage. (Thanks to Dark Kobold for making the TAS)
1. PSXjin outputs files at several resolutions. In order to join the files in Avisynth they need to have the same resolution/colorspace/framerate. So I made a source function to crop/pad them to match.
2. Most of the game is 240p, but the disc change scenes are 480i. Square seemed to intentionally mess up the interlacing here, but I think it looks awful, so I matched the same intensity fields.
avisource("M:\ff8_avi4\000_disc4-640x480.avi")
crop(31,38,580,406)
separatefields
#2 fields of Bright, 6 of medium, 4 of dim
#e e e e e e
#B;;.;.;B.;.;
# BB;;..
# 0769A5
selectevery(12, 0,7, 6,9, 10,5)
#pattern is always the same, but it doesn't always start on the same frame
weave
interleave(last,last)
#double frames to get back to 60Hz
#deleteframe/duplicateframe to match original frame count
3. PSXjin outputs A LOT of files. 178 avis for Disc 1. Avisynth can't usually open that many files at once, and typing them all in would be painful anyway. So I made a batch file that would take the name of a directory and splice together the avis into more managable pieces.
@ECHO OFF
SET maxsources=25
SET customsource="C:\Program Files (x86)\anrichan3.3\cs.avs"
SET counter=0
SET counterB=0
CD %~1
IF EXIST "spliced\" (
ECHO Spliced files already exist. OVERWRITING?!?
PAUSE
)
MKDIR "spliced\"
SETLOCAL EnableDelayedExpansion
FOR %%G IN ("*.avi") DO (
IF !counter! EQU 0 (
ECHO import^(%customsource%^) > "spliced\!counterB!.avs"
) ELSE (
ECHO \++\ >> "spliced\!counterB!.avs"
)
ECHO customsource^("%%~fG"^) >> "spliced\!counterB!.avs"
SET /A counter += 1
ECHO !counterB!.!counter! %%~fG
IF !counter! EQU %maxsources% (
SET /A counter = 0
SET /A counterB += 1
)
)
ECHO Encode Files?
PAUSE
CD "spliced"
SET vdub="C:\Program Files (x86)\anrichan3.3\VirtualDub-1.9.7\vdub.exe"
SET vcf="C:\Program Files (x86)\anrichan3.3\spliced.vcf"
FOR %%G IN ("*.avs") DO (
%vdub% "%%G" /i %vcf% "%%~nG.avi"
)
PAUSE
GOTO:EOF
Make the .vcf file in virtualdub, setup your encode setting, choose "Save processing settings... Crtl+S" from the file menu and add this line to the bottom.
VirtualDub.SaveAVI(VirtualDub.params[0]);
4. PSXjin outputs at a nonstandard framerate. (59.997?) I want to up this to 60, then resample the audio to keep it in synch.
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
natt wrote:
That's a pretty bad tradeoff
Yep, it is called placebo for a reason.
--preset veryslow --slow-firstpass --no-fast-pskip --bframes 16
would be the same as
--preset placebo --me umh --subme 10
but might convey the intent better.
Weakening the --preset seems odd to me.
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
It is easier to blur a sharp vid than sharpen a blurry vid, because blurring loses information which can't be regained.
What system are you using that you can't see pixels on your TV?
Break out an Atari and you'll see characteristically blocky groups of pixels.
Transitions from one color to the next were sharp (possibly a little noisy), but limited color palettes meant you never had the smooth color transitions that bilinearresize comes up with.
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
ledauphinbenoit wrote:
AviSynth does not support higher bit-depth by default and the hybrid script makes no mention of how to convert.
You don't need to convert anything. 10bit x264 will produce more efficient output because it is using higher precision for its internal calculations, even when using an 8bit source.
P.S. there is a missing newline in second_10bit444.bat
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
No, be so kind as to explain to me why it's a joke.
YV12 has chroma subsampling. Pointresize( *2, *2) before the color resampling will help preserve the information contained in the RGB original.
CRTs generally don't have square pixels. The NES renders a 256x224 color field which gets analogly spread onto a 4:3 screen. Stretching to 298x224 is a lossy process, but pointresize( *14, *12) isn't.
Youtube doesn't let you choose your upsizer when viewing fullscreen. Knowing that a truly pointresized option is availible at all resolutions is comforting to some. Others like to show off how shiny hqx/nedi makes things, while the rest of us gag.
Youtube doesn't always honor your resolution when it doesn't fall exactly on 240, and letting youtube decide how to handle resizing is pretty scary.
Pointresize( *14, *12) seems crazy, but with pristine sources and lossless encoding it compacts incredibly well. Lossy encodes will oddly be much bigger in most cases.
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
feos wrote:
The round turns into a square after processing.
You won't be able to do blinking and the round. TASblend would probably be a good compromise there, but it might mess up the motion in the rest of the scene. A version that turns just the blinking into transparency should be possible.
Baxter
Experienced Forum User, Published Author, Player
(65)
Joined: 4/21/2011
Posts: 232
Those seem to work ok.
A trim(1,0) was needed for the Turrican, I guess it is easier to overlay the larger sprite than try to cover it up.
GunstarTurrican
Being able to work on just the part of the scene that is flickering is nice (when it works).