Ghosting and artifacts?
 

ghosting and artifacts are still visibly clear , even though i increased the variance and max_details for both blockbusters line(0.5 ,0.7(variance)--8,10(max_details)) with script V4..although the walls(background) look pretty static now..is there anything else i could try??

check your source. does it have ghosting and artifacts? and the blockbuster line shouldn't be put up that high. Tell us what resolution your encoding at. mpg1/mpg2?

my source is a 2 disc xvid which i joined as one and now I am trying to make a kvcd out of it with 352x480 (MPEG1)resolution becoz my movie is like 2 hours 15 mins...and my source doesn't have ghost or artifacts..

btw source is at 608x320,23.976fps

if you're using the V4 script, the only thing i could recommend is enabling 3:2 pulldown in TMPGEnc. That should almost eliminate ghosting. and unless your source is extremely blocky, you could leave out the second blockbuster, and put the settings for the 1st one as:

Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.5,seed=1)

The problem is : what are you calling "Ghost" ?

A ghost image, or ghosting effect, is a near transparent copy of the image that is on the main image but moved by a few pixels.
There is no way that a ghost can appear with the filters we use !

DialHot...what u said is what i meant...i see a transparent "halo" over objects (humans,clocks,etc)...i'll try muhali3's recommendation and will get back to you ASAP

i did a couple of test using muhali3 recommendation and here are the results:-

i did this sample testing with a 1min21sec clip of an avi source where i feel the scenes in there should be fixed(walls,humans,etc --notclear)

The CQ for all my settings were at 70

1--script v4(544x480) -- size was 6984kb
2--script v4(544x480)(with 3:2 pulldown) -- size was 10315kb
3--script v4(352x480) -- size was 5323kb
4--script v4(480x480) -- size was 6412kb

The 1st one was my best with ghost scenes was about 85% removed (to my eyes)
The 2nd one was somehow jerky and there were "flying frames "around
The 3rd one was my worst with little or no improvements at all...ghost scenes were still there
The 4th was the second best with ghost scenes about 75-80% removed

What you are reporting is a complete nonsense. The filters are the same whatever the resolution so the results should be the same.

Can you upload a little sample somewhere ? I had to see this ghosting effect.

well i really do like to post a sample to show you... but I AM A COMPLETE NEWBIE...i just don't know where to begin...i know you'll be really MAD this time but please bare with me...can u give me directions??i already do have a lycos account but dont know how to upload

Just to be sure about something : you didn't check the "reduce ghost" feature in tmpgenc ? Did you ?

i didn't...must i check it ??

Not at all. I wanted to be sure you didn't because using it on a picutre that don't have a ghost default... CREATES a ghost during the encoding :-)

okey DialHot i got 3 images for you to see

http://www.digitalfaq.com/archives/error.gif

http://www.digitalfaq.com/archives/error.gif

http://www.digitalfaq.com/archives/error.gif


As you can see the 1st one is blocky and there still could be seen slight "halo""ghostly" image of the guy's left ear(this part i am not really complaining about) ...what should i do to remove the blocky part???

this is the script i used

LoadPlugin("C:\Filters\MPEG2Dec3.dll")
LoadPlugin("C:\Filters\grip.dll")
LoadPlugin("C:\Filters\undot.dll")
LoadPlugin("C:\Filters\blockbuster.dll")
LoadPlugin("C:\Filters\DctFilter.dll")
LoadPlugin ("C:\Filters\Convolution3DYV12.dll")
LoadPlugin ("C:\Filters\Sampler-2.5.dll")

AviSource("E:\movie.avi",false)
BlindPP(cpu=4)
Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.5,seed=1)
Convolution3D(1, 6, 12, 6, 8, 2.8, 0)
GripCrop(544, 480, overscan=3, source_anamorphic=false)
GripSize(resizer="LanczosResize")
Undot()
TemporalSoften(2,7,7,3,2)
DCTFilter(1,1,1,1,1,1,0.5,0)
#Blockbuster(method="noise",detail_min=1,detail_ma x=10,variance=0.3,seed=5623)
GripBorders()

CQ of 70

What you call "Ghost" is a mosquito noise effect ! For sure, there is nothing to do with a semi-transparent copy of the WHOLE image, moved by some pixels, that I describe in my previous post :evil:

If you have this despite of all the settings in the scripts that mean that your Divx is too bad (even if you do not notice the default visually, trust me, they are there). Forget it.

Crap in, crap out, there is nothing to do.

The same the for blocky walls : the script is able to reduce the effect, not to remove it as it is an internal problem of MPEG encoding.

i see...so is there a way where i can improve it (the wall scenes) or do i just have to bare with it

You already did all what you can do in the script.

You can try to use the feature "soften block noise" in the tab "Quantize matrix" of TMPGENC but it is not recommanded.
(it softens the blocks, but also the rest of the image ;-))

Dialhot, ... did you already do some tests on the "Block adaptative sharpening" and "High quality block filtering" parameters in Asharp???

http://www.media-video.com/contrib/c...ins/ASharp.htm


I tried a lot using diff. parameters but :arrow: :?: :arrow: :( = almost no difference.

... what are your results?

okey tried that...u're rite ...didn't help at all..overall image went soft...anyway thanx for ur help... :wink:

Never used any sharpening filter, including asharp.

(in fact I did some test but really do not like them, so can't tell if I saw a diff or not on the part you mention).

@the_ray
Well, my sugestion is - encode to 2 CD's.

@Abond
I would really like 1cd instead of 2cds...as one the members said
"if it ain't 1cd ,it ain't worth it" or something like that..sorry can't remember who or what he actualy quoted..

@DialHot
I would like a custom resize of 544x288 but when i run the script with TMPGe ,it gives me 544x416

AviSource("D:\movie.avi",false)
BlindPP(cpu=4)
Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.5,seed=1)
Convolution3D(1, 6, 12, 6, 8, 2.8, 0)
LanczosResize(544, 288, 0, 0, 608, 320)
Undot()
TemporalSoften(2,7,7,3,2)
DCTFilter(1,1,1,1,1,1,0.5,0)
#Blockbuster(method="noise",detail_min=1,detail_ma x=10,variance=0.7,seed=5623)
AddBorders(0, 64, 0, 64)

i had to do it manually with TMPGe..i know i am forgetting to do sumthing with the script so can you help me??ThanQ

288+64+64 won't result in 480! And therefore if you do it manually in Tmpgenc AFTERWARDS incl. these lines in your script it could end in a wrong aspect ratio when watching on Tv.

Therfore you calculated wrong. Did you use moviestacker or FitCD?
Where did you get these values?

actually it was 544x352 but i change it to 544x288 iwth moviestacker...the truth is i want something that looks 16:9 on my 4:3 TV ..so that i can't see any blockiness or such...544x480 almost fits the whole screen but i want something whoch is more or less 16:9

nevermind managed to figure it out...used 544x160 in the script to get 544x288 in tmpge...thanx anyway

Hi, the_ray,
I am afraid you will have wrong aspect ratio, if your standalone DVD player accept such a resolution at all.
The acceptable resolutions for standalone DVD players are 720x480, 704x480, 544x480, 526x480, 480x480, 352x480 and 352x240 for NTSC.
Your TV will stretch the frame to 4:3 despite which resolution you have in the mpg. Think a bit about it. Take in mind that the frame consist of the picture field and two black borders on top and bottom.

do u mean to say that even though the resolution of the the movie is still 544x480 , i can't make it to look like a 16:9 rather than a 4:3???

Of course you can ! But by adding black borders, not by doing like you do.

You told you used Moviestacker, so trust it ! It gives you all the correct seeting to have a perfect 4:3 picture that "looks like a 16:9".
(and you can even do a rela 16:9 picture if you have a widescreen TV set).

okey ..i see ..i think i am beginning to understand how this part works..regarding the black borders, do i add it at the beginning or at the end of the script and if so will the black borders crop out the movie or will "gripborder" take care of that part

Generally it's Croping -> Filtering -> Borders.

Some script do some filtering also before the croping but you never have any filter after the add of the borders.

If you use gripborders() you don't need to add them, gripborders will take care. Ofcourse you should use Gripcrop and gripsize lines as in the script you have posted before. BTW with that resolution I don't believe you will manage for 1CD.
Oops, I#m typing very slow :D

will this script do: -

LoadPlugin ("C:\Filters\MPEG2Dec3.dll")
LoadPlugin ("C:\Filters\blockbuster.dll")
LoadPlugin ("C:\Filters\UnDot.dll")
LoadPlugin ("C:\Filters\DctFilter.dll")
LoadPlugin ("C:\Filters\grip.dll")
LoadPlugin ("C:\Filters\Convolution3DYV12.dll")
LoadPlugin ("C:\Filters\Sampler-2.5.dll")

AviSource("D:\movie.avi",false)
BlindPP(cpu=4)
Blockbuster(method="noise",detail_min=1,detail_max =10,variance=0.5,seed=1)
Convolution3D(1, 6, 12, 6, 8, 2.8, 0)
GripCrop(544, 480, overscan=3 ,source_anamorphic=false)
GripSize(resizer="LanczosResize")
Undot()
TemporalSoften(2,7,7,3,2)
DCTFilter(1,1,1,1,1,1,0.5,0)
#Blockbuster(method="noise",detail_min=1,detail_ma x=10,variance=0.3,seed=5623)
AddBorders(32,64,32,64)

will it by any chance increase the CQ value??

Remove the AddBorders line from the script. As I wrote before gripborders will add automatically the right borders for you.

As you destination also will not be anamorphic YOU WILL obtain more CQ!
Cause the encoder doesn't have to encode that much active pixels cause of black borders using letterboxing! Thats why I recommend NOT to encode KVCDs as anamorphic if a prediction results in less CQ!


Well test your script by watching it in WMP or Zoomplayer, you will see the added borders ... don't worry if the image looks a bit squeezed at it's width cause you set 544 at width-resolution ... but you can determin a 4:3 output in Zoomplayer so you see the result as the same as it will come out an your Tv.

Or just for testing use (720,480, overscan ... etc) this will give you even mediaplayer used a "real-world-test" view.

;-)

BUT: If using Gripfit :arrow: get rid of the line "Addborders..." as already borders will be added by "GripBorders()"!!!!
(Uops .. refering to the last point Abond was faster ;-) )

but gripborders make the movie size larger than i want....i want the height of the movie to be smaller not bigger

Well in this case you should crop a part from the picture. You should yourself decide if it is worth for you as that means you will loose picture information. You should cut the heads or the legs decide yourself. If you want to have all the picture go for 2CD. There isn't other way around.

@ ray

I think its getting a bit confusing in here...

so decide what you want!:

1. Widescreen
http://www.digitalfaq.com/archives/error.gif

2. 16-9
http://www.digitalfaq.com/archives/error.gif

3. 4:3
http://www.digitalfaq.com/archives/error.gif

Green borders just show the overscan area!

www.incredible.de.tf/aspectratios.html
There you can find the explanations ... and therefor you can tell us what you REALLY want:

- 2.35:1 (Widescreen) (if your avi got widescreen proportions!)
- 1.85:1 / 1.78:1 (16:9 view)
- 1.33:1 (4:3 - full screen view)

??? ;-)

So if I understood it right Abond he wants a smaller active pixels size (like letterboxed 1.85:1 or 2.35:1) and by cropping you will perform the opposite ... a PAN-SCAN .... as explained in my site. ;-)
(Or did I understood you wrong? Confuuuuuuusing :D )

Hi, incredible
He has 1.9:1 (608x320 VGA) and he wants to make 2.35:1 :!:

@Abond
u mean to say that if i use AddBorders instead of GripBorders with the settings i put for AddBorders , the image will be cropped even more(as in heads becoming half-heads)

@incredible
I want the widescreen one not the 16:9(my mistake,i thought 16:9 was widescreen)..

sorry for misleading you all

@ Abond & Ray ... Thats very difficult cause ---- its almost impossible to do a Inverse PanScan cause the pixels on the sides are already cropped. :(
So if he got a 1.85:1 he should encode as letterboxed 1.85:1.

What Ray can do is performing an optical trick:

Using An "RESIZED" overscan by choosing a overscan value of 3 .... this will shrink the image within the picture BUT he has to watch out cause this will give 24px borders at the sides an dependending on the overscan of HIS TV set its possible that he could see the borders on the sides when watching the movie!

But just try ... Jellygoose for example got no problems by adding 24px borders at the sides. ;-)

So the image will come out like this:
http://www.digitalfaq.com/archives/error.gif

(Here in comparison again if you would encode WITHOUT resized overscan):
http://www.digitalfaq.com/archives/error.gif

Ray, ... you see the difference in the sample with resized Overscan above? :)
More "viewable" black borders seen on TV by keeping the active movie information! .... and you also will enjoy MORE COMPRESSION! :)


Keep in Mind that the green area is the overscan you won't see ... so by this you will receive a "viewable" average between 2.35:1 and 1.85:1 ;-)
The green areas will be black .. done by Gripborders() or in case on conventional resizing by using addborders or (if overlapping and NOT resizing) Letterbox().

I do encode ALL my 1.85:1 sources performing a RESIZED overscan!
This gives me a full compression advantage cause of less active pixels which have to be encoded ;-)

BTW: If using Moviestacker or FitCD ... Use lower right the option "Macroblock alignet" and NOT "centered" .... this will keep the 16x16 refered Macroblock Matrix of the image :arrow: also a plus for compression enhancement.