I respect your opinion but, how much time have you spent on your test ?
+ Reply to Thread
Results 31 to 60 of 328
thanks for new version.
Thanks for the new release, The despot/cleaning mode is excellent! i`ve never tried any AviSynth scripts before apart from those in XviD4PSP 5
and Film 9 works very well- expecially on dirt and scratches on motion, That said i have a query!
I imported a Pal Mpeg 2 file thats encoded interlaced but is progressive and i get severe colour banding even when all options are unchecked or turned off,Admittedly this clip is rather poor quality with lot of grain but i was curious why Film9 should have any effect when everything is unchecked etc,
Is this a colour space problem? I will try and upload a small orignal seqment of the clip, Its a old 1950`s B&W Clip but is still in copyright
so hope 30 secs should be okay?
Its also apparent that when everything is unchecked and disabled there appears to be some degraining going on that highlites the banding worse.
ps lordsmurf, I`ve used VReveal and its nothing like Film9, Please give it a go on some old film with scratches,its very good (colour banding issue aside Lol)
What does that mean - color banding on a black and white video?
I was working on a colour clip as i was posting so the word "Colour" was in my head !
The banding on the B&W clip affects A grey wall in the clip during a fade in, I`ll post a sample soon.
Here is the banding problem with my clip
As you see there are additional artfacts in the grain on the grey wall,this is with no cleaning applied or boxes checked.
Here is a sample of film9 cleaning,Notice the big spot almost removed on the arm, This is with the minimal + box checked
There is no way on earth "Vreveal" could fix that!
Here is the original Mpeg 2 Pal clip of 20 secs duration
Hope I`ve uploaded all this okay as I`m new here.
Here is that clip restored using my version of VideoFred's film restoration (both of these are documented over at doom9.org).
Of course once the clip is cleaned up, stabilized, and sharpened, the horizontal moiré interference lines (caused by IM distortion, adjacent channel, Kinescope artifacts, who knows?) become more prominent. It is possible that some of the amazing technology created by StainlessS over at doom9.org could be used to reduced those somewhat, but it will be a difficult task.
I too have been working on something similar, though it's main emphasis is extracting progressive frames from a film capture using a projector without a shutter running at around 6fps, though it should work with faster projections if your video camera can do 60p. It's a command line program that allows a lot of customization with regards to finessing the video for analysis and final output. You can find it at cinextractor.blogspot.com. It too is completely free.
Something I've found that works even better than the GetDups.dll is using the built-in Avisynth command Compare. You can compare adjacent frames with it and it will write out a text file providing various metric values of the differences between the frames. My program analyzes these values, selects the appropriate frames and write them out to another text file along with a new Avisynth script that reads the frames text file using StainlessS's FrameSel.dll and optionally opens this script using an encoder of your choice. I've gotten good results using this method and it's much faster than the GetDups.dll.
Let me know if you're interested in trying this method and I'll show you what needs to be done.
Depending Cassette option, it is true that a light DeGrain, and a light DeBlend is applied by default. Not in Film option.
Is this really penalizing ? Thank you to tell us your opinions.
I also grant that the Sharpness can be too strong with some tapes, so it is advisable not to use it.
Restore- chappy, thank you for this "wonderful" clip (too short) : THE SHADOWS, my favorite band (Marvin, Welch, Jet Harris and Tony), and of course CLIFF. In the first years !!!! SUPER . This affects me specifically !
John : We also follow the work of Stainless on Doom9, but it is only at the stage of the study. There is a lot of work, actually .
Khaver : Yes, I am very interested in your work and in particular by a potential competitor of Getdups who has its limits .
Thank you in advance for your advice.
Last edited by GillesH; 5th Feb 2015 at 10:03.
Decimate or MultiDecimate removes duplicate frames leaving unique frames only. If you were to use this on a video capture of a projection where the shutter has been removed from the projector, Decimate/Multidecimate would leave the frames where the film is being pulled down in the gate since they would be unique, along with 1 frame from the group of good frames while the film as at rest in the gate. The GetDups plugin compares adjacent frames and keeps the ones that are not unique (the frames while the film is at rest) then chooses the best one from the remaining duplicates and outputs it so you get a video clip containing only the frames of the original film. The GetDups plugin can also write out a metrics file showing the differences between the frames so you can have a program analyze this file and extract the good frames by writing a new file containing a list of those good frames and have StainlessS's FrameSel plugin read the frame list and extract those from your capture file.
My utility, Cine-XtractorCL, works similar to this but instead uses the built-in Avisynth function "Compare". In an Avisynth script I bring in my video capture clip and prepare it for "Compare" by duplicating it and trimming off the first frame. This gives me 2 clips to compare, the original clip, and a clip that's 1 frame off. The Compare function inputs these 2 "clips" and writes out a text file with various metric values comparing the 2 clips. My program then uses the values in this metrics file to choose the best frame (the one that's most likely the same as a frame from the film itself) and write out a text file containing those frames. This text file is then used by StainlessS's FrameSel plugin to output those frames to an encoder.
The problem with the GetDups plugin is it requires a lot of tweaking of threshold controls to get good results. It can also get messed up when there are very dark or black or white (or any solid color) sections. My utility accounts for this by finding an average cadence value (the number of frames between the good ones) and uses this value when there are dark or solid colored sections.
Thanks John for your work on this clip, I like your results as they retain more detail than my current methods, My workflow is convoluted and although I`m not a complete noob i have a lot to learn from you guys.
I had a go at trying to lessen those interference/telecine lines with this VirtualDub plug called "2D Comb" Then i ran it through QTGMC in progressive mode 1
Heres my restored version below
@ Gilles Re Film9
Its would be great if you could add a additional option to disable the initial light Degrain/Deband,this would make Film9 very useful for me, The banding on certain clips is the only reason i cant use Film9 for now,
Yes, you could use the metric file created by Multidecimate. The files it creates gives 1 metric number per frame. You then have to figure out a threshold number to use for selecting which frames are blurred from moving in the gate and which are at rest. Sometimes the video camera will capture a frame just as it starts to move or just before it comes to rest. This can mess up the metric number where the threshold number you chose won't work correctly. If you look at the output of the Compare function after feeding it with the 2 clips (original clip and the 1 frame trimmed clip) there are 5 metric numbers given for each frame comparison. By using all those numbers I've been able to select the correct frames pretty much without having to worry about a threshold number. You just have to make sure your captures have good exposure. If you need to to get good metric numbers for poorly exposed films, you can add an AutoLevels function before the Compare function at the expense of slightly slower processing.
Film9 accept not only progressive file. If you have problem with file format, you can send us a sample of your video so that we can test it.
In "FILM9/Info Clip", we use the same methods as MediaInfo .
Check with Gspot or MediaInfo if you have the same information.
Thank you for confirming .
Yes, in some cases, Sharpness is sometimes too strong default (especially in VHS) .
But, you can reduce or disable.
Certain indications, particularly in the screens of VirtualDub, are not easily translatable because integrated into the scripts.
All advice are welcome.
nice project and interesting approach, especially the preview you get thats nice. some small things:
* allow a "reset options" thats sets the sliders back to some sane defaults
* crop "automatic" is unclear as it enables the sliders - shoudl it read "active"?
* explain a bit what "attenuate" actually does
* "Refine" does pixelate the film, can't actually use it
Long term idea:
drop the "one script" style and allow to use e.g. your avs script and freds or johns - each script had to be annotated with a description for the input parameters for each filter componet used and generate a dynamic UI in a flow layout style based on the descriptions.
Each Filter could be activated or not and the avs script will be regenerated on each change.
so e.g. have a xml file
<filter "colors" description=" color adjustment" >
<input name="gamma" type="decimal" min="1.0" max="5.0" default="1.0" description="tooltip help gamma text">
<input name="blacklevel" type="integer" min="0" max="255" default="0" description="tooltip help blacklevel text">
...more params ....
OUTPUT=INPUT.coloradjust(gamma, blacklevel, ...);
<filter "deshake" description=" color adjustment" loadplugin="deshaker.dll" >
<input name="maxstabH" type="integer" min="1" max="50" default="1" description="tooltip help blub">
<input name="est_left" type="integer" min="0" max="255" default="0" description="tooltip help...">
<input name="abooleanparamater" type="boolean" default="false" description="tooltip help...">
.......more params ...
this woud allow to generate a avs script using the filters activatedd in the GUI with the parameters defined and run virtualdub on this avs.
this implies that every filter uses one input and generates one output and is chained together
i hope you get the idea,this would make a really powerfull tool!
Just a few answers:
- To reset the settings you must choose "Preset" / "Default"
- Yes the autocrop automatically AND it is possible to change slightly the values chosen by the autocrop.
- "Attenuate" allows you to view images on a TV without blowing colors. If you watch your video on a computer monitor, you can disable this feature.
- "Refine" ... ok, we take note
Thank you for your ideas ...
I`m still having issues with banding , The dust/spot removal modes are excellent but i get this horrible banding on the grain,
I would like to just remove spots without affecting the grain too much as i want to use other noise removal methods like "Neat" after the spots are fixed,
Here`s a example pic
Heres a original clip of the source, (720X480 29fps)
I applied IVTC to the video in the picture to remove the dups and use a 23,976 HuffYUV as input to Film9.
Here is a draft version of FILM9 : https://www.dropbox.com/s/8qvv5a927lvoo41/Film9.zip?dl=0
This version disables (or not) LEVELS. Which can cause this "banding" effect on Black/White film.
Tell us whether this is enough.
Thank you Gilles and Roland for the new version, Its much better and has reduced the banding considerably- but not completely,
I think the banding is a side-effect of the spot/dirt remover as the banding disapears when the "Cleaning" box is disabled
Because i notice there is a some slight degraining applied with the spot remover?
Please dont take my remarks badly as i appreciate what you are doing. This latest Film9 version is very useable and i notice the banding only on very bad material
Could you explain the "Cleaning" functions `+ + S` ( I asssume the "S" is for Scratches?)
It will be difficult to delete everything because any remaining filter will, inevitably, impact the image.
For cleaning , there are 3 modes:
- Cleaning + : single image analysis
- Cleaning + : there is a double analysis of the image for better quality , but it's slower.
- Cleaning S : it is, too, another double image analysis but to treat clearer spots.
The differences between modes are sometimes very subjective. It also depends on the quality of the movie.
Would it be possible to add some sort of re-grain option ?
Or maybe a option to take a sample of the original grain and use them as presets, Then have a slider to vary the amount of applied grain after restoration.
Also is it possible to add my own selection of Virtualdub filters as Film9 processes and saves the clip?
Thank you for your ideas, but :
- It is true that the grain addition can sometimes compensate certain defects.
But it is likely that it will be easier to do this job in a NLE.
- For VirtualDub filters, we can not add them to FILM9 because the definition of the Process is declared in the Code (JOBS file).
Sorry for these limitations.
And thank you for your attention to this software.
John responded well and I answer as Simon .
All software developed with Visual Studio require the NET.Framework .
And it is our choice !
Video computers should be lean, not full of unnecessary shit. That's why.
There can also be unforeseen consequences of installing various .NET Framework versions, and they can screw up far more important function of the video system. That's why.
I don't want to fubar my stuff just to try a whizbang new tool.
(FYI: The animosity is towards smartass simonb, not you johnmeyer.)
After moving to a safe system, I tried this. It's mostly useless, as far as I can see. I ran into foreign-language error messages, and the whole "you must setup capture first" was really stupid. It's a halfass attempt to be all-in-one software, rather than the "film restoration" tool that it claims to be.
The ONLY thing that I was interested in trying was the NR for film noise. But I couldn't get that far.
It's rather sad when Avisynth is actually easier.