Color correction (or grading?)...

[takearushfan]

Okay, just by asking this you'll know that I have no background in this field so any/all info is greatly appreciated.

For a long time now I've wondered what exactly color correction is and how it works. I read briefly that in the case of video it may be referred to as color grading. Nonetheless, from hereon in I'll use the term color correction but know that I'm referring to video.

It's my assumption that color correction is just as it sounds. Let's use the analogy of a DVD being color corrected from a VHS source, or a Blu-ray being corrected from a DVD, etc. How does whoever does this correcting know what the remastered colors are supposed to look like, as opposed to the old "wrong" colors? Do they go back to the theatrical print and somehow work with that?

I've read about many applications (with hefty prices) that are designed for such purposes. Some of these seem to imply that the average user can due such correction. I don't quite understand how, unless you have the original print to work with...?

Naturally the average user wouldn't have such a source, equipment, the know-how and the plethora of other enormously expensive hardware necessary.

So, what's the deal here?

Thanks,

Justin

[poisondeathray]

Well there is a difference between color correction and color grading

Color correction refers to balanced colors and tonal range, like proper white and black level, neutral hues.

But most hollywood material is graded. Theatrical releases are very rarely color neutral or balanced, especially in modern releases. This is purposely done by the colorist and director to help tell the story or create atmosphere

So what is it "supposed" to look like? I don' t know . This is very subjective topic. Is it what the original director envisioned it to look like ? Even different studio releases and transfers of the same material can look very different. Sometimes it's done on purpose, sometimes its a botched job.

Promotional material like posters and commercials often looks slightly different from the theatrical or release on blu-ray. Often different people work on these materials (graphic designers on print material, but colorists on the DI), the print material is usually more neutral

[takearushfan]

Hm. So unless the colorists are working right beside the director, a colorist is (kind of) "winging it"?

[poisondeathray]

The director always has to sign off (at least is supposed to)

But there have been some pretty bad "remasters" that make some people think the director was too lazy or sleeping...

Usually there are multiple revisions and back and forth - and even if the director isn't sitting right there, the colorist will send off a rough grade to get feedback and comments

[takearushfan]

I read reviews saying that Suspiria (on Blu-ray), which is not yet available in the states (GRRR!), has some color issues. It now appears that one of the characters has hands that are FAR too white compared to past releases. However, they said that the coloring of the school is more pink now than how it used to be, red. It's said that the color of the school is actually more genuine to how Dario Argento wanted it to be in the first place. So I guess they kind of screwed up the teacher's hands but properly corrected the school.

[sanlyn]

I often see your question in this and other forums. I asked that question myself. But I was surprised to find tutorials on every subject under the sun except one: what is color correction and how is it done?

poisondeathray has a good general reply that explains some basic terms. I'd add that "color grading" also involves "color timing". If you read some of those rolling credits at the end of some films you'll see a credit for "color timing". This is a form of color grading referring to color correction in video sequences that makes every camera shot look as if they were made by the same equipment at the same time. Example: a guy walks into a scene wearing what looks like a dark gray suit, blue tie and white shirt. In the next camera angle of the same scene, the suit looks dark blue and the tie is taking on a purple hue. Or a gal in one shot is a blonde, but two shots later in the same sequence she's turning into a redhead, then she's blonde again. Special lighting and other effects aside, clothing, hair, and skin color shouldn't change every few seconds. It annoys most viewers, and it's a sign of sloppy work.

Let's use the analogy of a DVD being color corrected from a VHS source, or a Blu-ray being corrected from a DVD, etc. How does whoever does this correcting know what the remastered colors are supposed to look like, as opposed to the old "wrong" colors? Do they go back to the theatrical print and somehow work with that?

Sometimes it's done that way. Sometimes it's a matter of correcting the source and the final product alike. You'd be surprised how many original prints have color issues, especially if they're aged or damaged. Video restoration jobs like How To Marry A Millionaire and Rear Window are cases in point: the original prints and negatives had one or more badly faded color layers.

I've read about many applications (with hefty prices) that are designed for such purposes. Some of these seem to imply that the average user can due such correction. I don't quite understand how, unless you have the original print to work with...?

You don't need the original print - good to have it, though, to see what the creators really intended. The tools you pay for with $800 software and $1000 plugins are quite sophisticated, with many automated features, displays, vectorscopes, comparison panes, buttons for saturation, contrast and whatnot, and buttons that magically change everything blue into red, etc. Few people actually use those automated features; rather, they make their own analyses and decisions.

Expensive tools are nice, but Avisynth and VirtualDub have freebies that look and work in similar ways to the big guys: histograms, vectorscopes, waveform displays, gradation curves, YUV and RGB color filters, multifunction and multi-range plugins, etc. Many complain that they have to read some documentation and learn how to use these tools. But the $800 guys require learning as well, often for fees that surpass the cost of the products.

Naturally the average user wouldn't have such a source, equipment, the know-how and the plethora of other enormously expensive hardware necessary.

See above. No, you don't have the fancy features but Avisynth/VirtualDub tools work the same way because the basic principles of correction are the same. That goes for video as well as still images, which is why the vast number of free Photoshop tutorials on the internet are such bargains.

The know-how you mention comes from several sources. The primary source is developing a good color eye by observing the world around you. Special effects aside, fundamentally "correct" color should look "real", or what people like to call "convincing". Green skin tones and purple shadows in hair are neither convincing nor pleasant to watch.

Learning how to make those corrections involves learning about the components of color. We all know what Red "looks like", but in RGB terms a full 100% red has pixels with RGB values of R/255-G/000-B/000 (that is, this pure red is all red with no green or blue in it). The primary colors are Red Green Blue, but you'd have to know the secondaries as well: Magenta (Red+Blue), Cyan (Blue+Green), Yellow (Green+Red).

You also learn that three of the most important and useful "colors" are Black, Gray, and White. Each of these colors has equal amounts of three primary colors:
Black = RGB 000-000-000
Gray = RGB 128-128-128 (aka "middle gray")
White = RGB 255-255-255
You might wonder why photogs, cinematographers, and colorists work with grayscale ramps and patterns. They pften talk about "setting" the white point, black point, and gray point. That's because an image that can accurately represent those hues as true Black, White and various shades of Gray indicate that all three colors are properly balanced. When those three colors are correct, the other colors (including skin tones) fall into place.

The third area of learning involves using the tools and filters. Example: hobbyists and pros alike use pixel samplers -- a tool that tells you what the real color values of pixels are. If whites look too yellow, pixel readers help tell you why: is there too much green and red, or not enough blue? Examining other elements in the image will reveal more problems that help you analyze and correct imbalances. Various histograms and graphical 'scopes, etc., are free tools that compete with the big guys. Most of these free tools come with documentation and samples.

And don't be surprised if many videos defy complete correction. You do what you can.

Yeah, it's a learning game. You learn by watching and doing.

[poisondeathray]

I read reviews saying that Suspiria (on Blu-ray), which is not yet available in the states (GRRR!), has some color issues. It now appears that one of the characters has hands that are FAR too white compared to past releases. However, they said that the coloring of the school is more pink now than how it used to be, red. It's said that the color of the school is actually more genuine to how Dario Argento wanted it to be in the first place. So I guess they kind of screwed up the teacher's hands but properly corrected the school.

I think that's the key - it's a very subjective topic.

Color correction is supposed to yield balanced tonal range, neutral and technically proper "textbook" results, but often grading can lead to "not textbook" results - and as said earlier , often this is done on purpose. Common examples are crushed blacks, oversaturated looks , "filmic green" looks . When other people are involved, sometimes they alter the "look" so it goes against what the original filmmaker or director wanted. If it's an older film "remaster", sometimes the original team isn't available or director has passed. "Color correction" is easier, because it's almost a textbook formula. Grading is much more subjective - you have no idea of what the director wanted it to look like

And as mentioned above, sometimes different studios using the exact same master and DI come out with different colors. WTF? Which one is the "intended" version? Sometimes it's a complete screw up (although no one will admit to that)

It's a similar discussion with "film grain" (many productions add grain, arguably overdo it in some cases).

But sometimes things are clearly out of whack. You can't tell me Scorsese signed off on "Gangs of New York" remaster . Of course this is a relatively newer film, and he's very much alive






You can see more pics and discussion here
http://www.avsforum.com/t/1045420/gangs-of-new-york-comparison-with-remastered-version-update-pix


Sanlyn is right, you can do most color correction and grading with commonly available software, even free ones like vdub and avisynth filters . The difference between the expensive tools and the really expensive tools (e.g. Davinci Resolve workstation), is that the latter provides a more streamlined workflow. The controls are a lot better. The Resolve hardware makes things easier to do and faster than just the Resolve software version. They have better motion tracking ("power windows") and secondary CC capabilities , and importantly support for LUTs which are required for calibration.

For example , the teachers hands in your Suspiria example would have been motion tracked and secondary CC'ed had they wanted to . I would say >90% of feature films and TV productions are graded on Resolve (the rest are in-house software) . There was actually a free software beta version available "Davinci Resolve Lite" (limited to 2 nodes) a while back, it might still be available to play with

[sanlyn]

Poisondeathray makes some excellent points. Pricey software can do things that with free tools are possible (some of it, anyway) with free tools, but extremely difficult to accomplish. And the learning curve with sophisticated tools is something else: I've had AfterEffects Pro for 7 months -- I'm still learning details of the interface, forget about motion tracking!

Still, we poverty-stricken hobbyists can make improvements. And as pdr says, everyone makes decisions about the look they want. There are both subjective and objective considerations. Objectively, you'd want a video to comply with accepted level and color standards; A video that's blindingly bright or has crushed darks gives encoders a rough time and will look worse on TV. In the two images above, the top image is too blue and has clipped brights; the bottom image is too green and dim. Doorways and some shadows, even though dark, look foggy. As an example of how histograms are used to analyze level and color problems, here are 4 histograms of the top (bluish) image. I used Photoshop histograms here (in this case, they are easier to read), but you can ascertain the same problems with Avisynth or VirtualDub tools:

Top image histograms: RGB (overall luma), + Red + Green + Blue. At the right side of the histoghram, note howm the graph smashes against the right-hand wall with RGB, Green, and Blue clipped:

[Attachment 13075 - Click to enlarge]

Depending on the impact or impression you want to make with this scene, level and color corrections make a difference. Below is a quickie using VirtualDub's garadation curves plugin. Here, I chose to target natural looking colors under an overcast sky. You might like to see the scene another way:

[Attachment 13076 - Click to enlarge]

Keep in mind I'm working here with captures, not with the source. That in itself has its limitations.

[poisondeathray]

They should have graded the snow yellow

[sanlyn]

Yes, I don't know where that purple is from. But since the samples aren't from the source, I made quickies to show the "free" possibilities. The discoloration is from a Blue peak around RGB 220. But I agree, I'll bet that purple in the original (or in real life) would be yellowy. Material like this is good study fodder for learning AfterEffects and Color Finesse. I paid enough for it, I may well learn to use it.

[poisondeathray]

That was a joke.. "yellow snow" .. get it ?

[sanlyn]

Razzing me again, poisondeathray. But the purple, well, I don't think the source had that problem. I would have first corrected in YUV to lower blue contrast. But I think you joked "correctly": those stains should be yellowy, or even grayish. Looking at the greenie version posted earlier, those same snow areas look incongruous. But that's another lesson to be learned: watch YUV colors and levels before proceeding.

[takearushfan]

Now, THIS is the kind of generous feedback and enormous info I love... GREATLY appreciated!

Here's one for you though. What about issues with contrast, brightness and such? I've noticed that if I'm careful, using the less is more theory, I can adjust levels somewhat well; somewhat. If there are scenes that are painfully dark (I'm assuming print degradation), you can't raise things too high or else you begin to see major blocking.

Cleaning is another bitch (pardon) ... I like Neat Video BUT, again, I've found that less is more. I've learned from much experimentation that overdoing it will turn characters into clay; too much smoothing. Someone's wrinkles, armhair, etc. shouldn't be eliminated. And grain, another PITA... I have to make sure not to remove too much grain. Some SHOULD be there, IMHO.

If you really think about it, when you see a film in the theater, even at the current date, it doesn't look stellar. You'll see grain and even artifacts from the print being played repeatedly.

I just hate how some masters are preserved so beautifully, ie. Tales from the Crypt (1972); can't believe how pristine it looks. Then you see When a Stranger Calls. What in the living hell were they thinking in releasing... eh- you know me well enough from other posts that I've already "went there" countless times. Sorry *shaking my head*

Yes, I rant and am long-winded why I get passionate and very interested in these topics. I apologize if I bore the hell out of you all :/

[poisondeathray]

Here's one for you though. What about issues with contrast, brightness and such? I've noticed that if I'm careful, using the less is more theory, I can adjust levels somewhat well; somewhat. If there are scenes that are painfully dark (I'm assuming print degradation), you can't raise things too high or else you begin to see major blocking.

The difference is you're attempting this on a heavily compressed DVD . Consumers typically don't have access to materials "higher up the food chain"

The intermediates used for celluloid film scans are usually 2k - 4k 16bit, full RGB color 4:4:4 - no subsampling, no compression artifacts

Your DVD is full of compression artifacts (yes, even the most pristine DVD is full of artifacts to the colorist or compressionists eye - mpeg2 is just a crappy compression), 720x480 Y' (the chroma resolution is only 360x240) , 8bit (the RGB values range only from 0-255 or 16-235 for legalized Y' and 16-240 for CbCr, but 16bit masters range from 0-65536 - that's way more accuracy and less banding)

The difference in grading and leeway you have is like night and day, even moreso with todays digital cinema (not celluloid, I'm talking about modern digital cameras like red epic , arri alexa and sony f65 ) . The only thing that celluloid had over previous generations of digital cinema cameras is more latitude. But this newest crop is very close and exceeds film in many other other catergories, certainly for overall production costs and workflow

For old films, it depends on what is still available, and the condition of it . Often it's the poor condition and deterioration that is the limiting factor , regardless of what type of intermedate or processing is available . Then you have those massive restoration processes (depending on the budget). For something like Jaws, it looks like they put in a good effort - you can view the story on youtube and various sites if you're interested in the process

Cleaning is another bitch (pardon) ... I like Neat Video BUT, again, I've found that less is more. I've learned from much experimentation that overdoing it will turn characters into clay; too much smoothing. Someone's wrinkles, armhair, etc. shouldn't be eliminated. And grain, another PITA... I have to make sure not to remove too much grain. Some SHOULD be there, IMHO.

If you really think about it, when you see a film in the theater, even at the current date, it doesn't look stellar. You'll see grain and even artifacts from the print being played repeatedly.

It depends what you're looking at and how you are judging it. You're obviously a big fan and critical, but your average viewer is usually more interested in content and story

Film grain is one of those things that is really subjective . Some love it , some hate it. You can look at the numerous other threads discussing this. Part of the reason compressionists (and directors) add grain (to existing film grain) is to hide defects , compression issues, and 8bit issues (banding)

If you've ever had a chance to see real 4K material on a 4K projector - you will be floored .

[takearushfan]

I did see the video on the restoration of Jaws and it's no wonder it took a while to release while so many were demanding it. I found Spielberg's comment on how the original print was "pretty lousy" to be humorous, yet they worked their *sses off to remaster it. Honestly, I don't own the Blu-ray of Jaws so I don't know how it looks, but I'm sure it looks great.
I also watched a video on the restoration of The Wizard of Oz. Wow! There was SO MUCH work involved, but this is one of the biggest classics of all time. Again though, I haven't seen it on BD.
In regard to WaSC, I understand that casual/first time viewers would be more interested in the story than being picky of visual/audio details. But, really... that shot with the massive scratch marks. Come on... ANY viewer would find that at least a bit unusual. Personally, it REALLY p*ssed me off. I don't think I'm overreacting here. You do know of what shot I mean, right? I've posted it before.

[takearushfan]

In the rare case that you don't know the shot, it's attached. I'd love to repair it myself but have no conceivable idea of a practical way to do so. IMHO, if they cared, Sony would've borrowed a portion that isn't disastrous from another print.

[poisondeathray]

In the rare case that you don't know the shot, it's attached. I'd love to repair it myself but have no conceivable idea of a practical way to do so. IMHO, if they cared, Sony would've borrowed a portion that isn't disastrous from another print.

Yeah, that' s pretty bad. It's obvious enough that your average viewer will definitely notice. There are also spots that they neglected to remove

But how do you know that there is another print out there doesn't have this ? (or did you see another cleaner version) ?

There is also a return on investment issue here; the "bigger" or more popular the movie, the more likely they'll spend the effort and $$ to fixing it up properly




Depends on your definition of "practical way" of reparing it and how much effort you want to put in.

What's normally done is a patch method . They fix up a few critical spaced out frames (in photoshop or restoration software) then use those to "paste" into other frames . Usually motion tracking is used to make things go faster instead of doing it all manually .

If you can make use of original material that is already there - that's better and faster. For example , about 3.5 seconds in there is a transient scratch over the man's left eye . Transient means you could take the adjacent "good" frames and motion track them to cover up the scratch . This saves a lot of manual work that you would have needed to do in photoshop or similar programs . But if there is a lot of motion, then usually there is not much information you can salvage from adjacent frames (the angle and the background that you're trying to copy from is different). In areas where the scratch is over no detail (like black background), that's obviously easy to fix. These are reasons why persistent defects like typical scraches take much more effort, but transient defects like spots are much easier to fix (in fact there are many "automatic" filters in avisynth and various programs that can do this)

This is usually done on the "clean" master (or clean as possible), otherwise the noise and grain pattern won't match up and there will be an obvious mismatch. Then you would have to degrain and regrain. Specialized compositing FX software have emulated film stocks (different film stocks have different grain characteristics - if you know the # the film was shot on e.g. Kodak 500, you can match up the grain characteristics more easily in post)

A lot of this you can do in consumer/prosumer software like photoshop, after effects. You can combine it will free software like avisynth , vdub, gimp - it just depends on how much effort you want to put in

[takearushfan]

In the rare case that you don't know the shot, it's attached. I'd love to repair it myself but have no conceivable idea of a practical way to do so. IMHO, if they cared, Sony would've borrowed a portion that isn't disastrous from another print.

Yeah, that' s pretty bad. It's obvious enough that your average viewer will definitely notice. There are also spots that they neglected to remove

But how do you know that there is another print out there doesn't have this ? (or did you see another cleaner version) ?

There is also a return on investment issue here; the "bigger" or more popular the movie, the more likely they'll spend the effort and $$ to fixing it up properly




Depends on your definition of "practical way" of reparing it and how much effort you want to put in.

What's normally done is a patch method . They fix up a few critical spaced out frames (in photoshop or restoration software) then use those to "paste" into other frames . Usually motion tracking is used to make things go faster instead of doing it all manually .

If you can make use of original material that is already there - that's better and faster. For example , about 3.5 seconds in there is a transient scratch over the man's left eye . Transient means you could take the adjacent "good" frames and motion track them to cover up the scratch . This saves a lot of manual work that you would have needed to do in photoshop or similar programs . But if there is a lot of motion, then usually there is not much information you can salvage from adjacent frames (the angle and the background that you're trying to copy from is different). In areas where the scratch is over no detail (like black background), that's obviously easy to fix. These are reasons why persistent defects like typical scraches take much more effort, but transient defects like spots are much easier to fix (in fact there are many "automatic" filters in avisynth and various programs that can do this)

This is usually done on the "clean" master (or clean as possible), otherwise the noise and grain pattern won't match up and there will be an obvious mismatch. Then you would have to degrain and regrain. Specialized compositing FX software have emulated film stocks (different film stocks have different grain characteristics - if you know the # the film was shot on e.g. Kodak 500, you can match up the grain characteristics more easily in post)

A lot of this you can do in consumer/prosumer software like photoshop, after effects. You can combine it will free software like avisynth , vdub, gimp - it just depends on how much effort you want to put in

The only other version I've seen was on VHS... back in '91. So, not only would I not likely have noticed it as much on crappy equipment/VHS but this was also too long ago for me to recall the flaws :/
I had just assumed that there would be another print that didn't have the scratching. I was told from some other poster that it looks like it really is the original print that was used, unfortunately damaged. How he knows that for certain is beyond me, but okay ... the problem is that if what we're seeing are heavy scratches and this IS the original print, then wouldn't it be useless to seek out another print, as all others would just be "based on" the quality of the original? Ugh, there are many potential factors involved but, as you mentioned, none of them will come into play, as the demand just isn't there. This isn't Jaws, The Exorcist, The Wizard of Oz, A Nightmare on Elm Street, yadda... this is simply not worth Sony's investment, from their POV.

As for repairing it myself. I'll need to reread what you posted a couple times after I have some coffee

[takearushfan]

The scratch on his face originally appeared in shot 84. I know it's hardly much work to do all of 4 frames, but I think I get what you're saying about the (for lack of a nicer term) "copy/paste" technique. That's really all I did here. Like you were describing, the problem does arise when even if the shot is a freeze frame, the grain begins to change too much.

[poisondeathray]

I had just assumed that there would be another print that didn't have the scratching. I was told from some other poster that it looks like it really is the original print that was used, unfortunately damaged. How he knows that for certain is beyond me, but okay ... the problem is that if what we're seeing are heavy scratches and this IS the original print, then wouldn't it be useless to seek out another print, as all others would just be "based on" the quality of the original?

This is probably the case, but you can't know for certain just from that information alone. Even the presence of a "non scratched" distributed version doesn't necessarily indicate that a "non scratched" print is available. For example, one studio might have a "fixed" DI version (unlikely for this movie...) . I'd assume that you were in contact with fans of the film and someone in that group would know or comment on a non scratched version if it existed

As for repairing it myself. I'll need to reread what you posted a couple times after I have some coffee

How much more footage is scratched in this version of the movie that you have? Does it occur through the entire movie or is it just this shot? I had a closer look and there is nothing that requires huge amount of work. The distribution of scratches are actually placed fairly favorably (meaning you can take data from adjacent frames to "fix" up the scratches) - at least in the sample you provided. ie. there not too much extra work needed in photoshop with paintint - it's just masking and positioning (or motion tracking if you have that)

[poisondeathray]

The scratch on his face originally appeared in shot 84. I know it's hardly much work to do all of 4 frames, but I think I get what you're saying about the (for lack of a nicer term) "copy/paste" technique. That's really all I did here. Like you were describing, the problem does arise when even if the shot is a freeze frame, the grain begins to change too much.

Yes that's the general idea. But do you notice the patch "drifts" a bit ? . It doesn't "stick" on his face. e.g. You can see his eyebrow come apart . It probably won't be noticable when viewed normally, but this is where motion tracking comes in - Instead of manually positioning the patch, you "link" it up (it's called "parenting" in the VFX world) to the motion tracking data

Matching the grain and noise pattern is part of the art of compositing - it's harder to do without VFX type software ike after effects or nuke. There are also grain plugins that sort of work in reverse to how neat video works - similarly you sample a grain area but instead of removing the grain , it tries to emulate and reproduce the grain .

On this particular shot, it's actually a static zoom of a still to begin with. The noise is compression noise, not grain until frame 107 (IVTC'ed numbers) . After that frame, you notice the grain start to dance - thats the real grain and where actual live action starts

On this shot it looks like the other scratches you can do mostly with frame offsets & the "copy & paste" patch technique - sometimes it's not possible it depends on the movement of the camera and what is "under the scratch" on neighboring frames . When this happens you have no choice but to paint & reconstruct in photoshop

The other "gotcha" that might give you problems is a slight luminance flicker . When patching in from different frames, the patch might be slightly discolored or have different levels and look mismatched - again that is part of the art of compositing . You make adjustments to luminance and color, keyframing the changes - it's harder to do without using at least editing software with a timeline

A tip that many people use is to temporarily enhance the scratches and other spots, so you see them more clearly (especially on a darker scene like this) - you can do this by increasing the gamma for example . This also helps validate your repair - if it's seamless with the "enhanced defect" version then there is even lower chance that the repair will be visible when you return it back to normal

[takearushfan]

I had just assumed that there would be another print that didn't have the scratching. I was told from some other poster that it looks like it really is the original print that was used, unfortunately damaged. How he knows that for certain is beyond me, but okay ... the problem is that if what we're seeing are heavy scratches and this IS the original print, then wouldn't it be useless to seek out another print, as all others would just be "based on" the quality of the original?

This is probably the case, but you can't know for certain just from that information alone. Even the presence of a "non scratched" distributed version doesn't necessarily indicate that a "non scratched" print is available. For example, one studio might have a "fixed" DI version (unlikely for this movie...) . I'd assume that you were in contact with fans of the film and someone in that group would know or comment on a non scratched version if it existed

As for repairing it myself. I'll need to reread what you posted a couple times after I have some coffee

How much more footage is scratched in this version of the movie that you have? Does it occur through the entire movie or is it just this shot? I had a closer look and there is nothing that requires huge amount of work. The distribution of scratches are actually placed fairly favorably (meaning you can take data from adjacent frames to "fix" up the scratches) - at least in the sample you provided. ie. there not too much extra work needed in photoshop with paintint - it's just masking and positioning (or motion tracking if you have that)

No, this is hardly the only scratch (and many other defects) in the film, but to be fair, this is the only portion that is what I feel to be too extreme/unacceptable.
There are many times when I'm not even sure what a particular artifact is caused by. I noticed some shots in this film that had the same artifact as when I viewed parts of The Day After (made for TV, 1983, 35mm). They're rarely a single "dot" but rather a more random "hair-like" look, but it's more like a blueish splotch.
Good grief... I found a stunning, beautiful shot that not only shows the blueish thing I don't understand but many other defects, all the blink of an eye.
P.S. - Sorry for the file size but naturally it's best to keep compression low when exchanging these details.

[poisondeathray]

No, this is hardly the only scratch (and many other defects) in the film, but to be fair, this is the only portion that is what I feel to be too extreme/unacceptable.
There are many times when I'm not even sure what a particular artifact is caused by. I noticed some shots in this film that had the same artifact as when I viewed parts of The Day After (made for TV, 1983, 35mm). They're rarely a single "dot" but rather a more random "hair-like" look, but it's more like a blueish splotch.
Good grief... I found a stunning, beautiful shot that not only shows the blueish thing I don't understand but many other defects, all the blink of an eye.
P.S. - Sorry for the file size but naturally it's best to keep compression low when exchanging these details.

Just a note for future - when uploading uncompressed video, it's best to use 7zip or some compression utility e.g. winrar, winzip etc.... It will reduce the filesize substantially (to about 1/3 of original size in your example) . Also uncompressed audio, you should use 7zip as well on the .wav files




Spots , dots, hairs, various defects etc... - there are multiple causes, but that's probably not what you're interested in. You're probably more interested in what can be done to fix them

When defects occur in single frames, that's where the "automatic" avisynth filters can work well, like removedirtmc and dozens of others like removespots etc....They look at differences between frames (so if a spot or hair occurs only in a single frame, it will be "detected" , but if it lingers in a few frames like a persistent scratch it won't be).

I would combine them with masks so you limit the repairs (otherwise other wanted objects can go missing like in this video, the ice falling might "disappear" and inadvertently get removed) . What you do is place a heavy filtered version on 1 layer and a non or lightly filtered version on the other and just draw or circle masks where the spots/defects are, or the reverse, use a blocking mask where you don't want to filter like a box around the falling ice chips. But this technique will save you hours/days/weeks of work .




You should also take care when converting to RGB, uncompressed RGB or still images . In this shot there looks to be clipping in the bright area of the fridge - the original video might have usable overshoots and salvagable detail that were thrown away by the Rec601 conversion to RGB

[takearushfan]

Note taken about the zipping advice, though I may not recall it when it comes time to upload due to force of bad habit

Before addressing your previous post, I'd hate to do this to you but would like some assistance. Actually this is due to reading your last paragraph.

Can you help me to take this...

General
Complete name: C:\VTS_01_1.VOB
Format: MPEG-PS
Overall bit rate mode: Variable
Overall bit rate: 5 556 Kbps

Video
ID: 224 (0xE0)
Format: MPEG Video
Format version: Version 2
Format profile: Main@Main
Format settings, BVOP: Yes
Format settings, Matrix: Custom
Bit rate mode: Variable
Bit rate: 5 445 Kbps
Maximum bit rate: 9 800 Kbps
Width: 720 pixels
Height: 480 pixels
Display aspect ratio: 16:9
Frame rate: 23.976 fps
Standard: NTSC
Color space: YUV
Chroma subsampling: 4:2:0
Bit depth: 8 bits
Scan type: Progressive
Scan order: 2:3 Pulldown
Bits/(Pixel*Frame): 0.657

...

and determine which scripting/filters (via Avisynth) would maintain the most accurate/preserved output? Yes, a bit out of the blue and tedious. This is why I said I hate to ask it of you

[takearushfan]

I never posted any image of my blue artifact question, found in WaSC and my yellow color question, for in The Day After.

Though if I were to look back I'm sure you've already addressed this, please remind me of what (perhaps during the telecine process?) cause the blue/yellow artifacts.

Yes, I blew up the images, roughly "pointed" to the artifacts, etc. I didn't care about preservation as you can get the point even if I mangle the images.

[jagabo]

I think those blemishes are from scratches or goobers on the individual red, green, or blue negatives, before the positive print was made.

[takearushfan]

I think those blemishes are from scratches or goobers on the individual red, green, or blue negatives, before the positive print was made.

That's what I was leaning towards, but since I'm as far from an expert as possible I figured I'd ask anyway...

When I was watching the YouTube video of the restoration process used for the BD of The Wizard of Oz, I thought it was a brilliant concept that they used which (supposedly, according to them) they thought up for the first time; I'll take that they discovered it with a grain of salt ...

They noticed that when trying to remove some artifacts, vital elements would be eliminated, ie. her shoes! Not good, so they examined each negative carefully and made certain that if a flaw exists on only one negative, then obviously that's a true issue to correct. If it exists on all three, it should remain. It seems so logical that I can't imagine them really being the first to think of it, but you never know.

Between TWoO and Jaws, it's apparent that (if the demand is there) an unfathomable amount of work goes into the restoration process.

It's too bad that if you're a fan of any other film but the demand is NOWHERE near what a studio finds worthwhile, the consumer is left with shite. This of course leads to people like myself probing everyone on how the average Joe can work on these things

BTW, thanks for your reply

[themaster1]

A good reading:
www.history.org/foundation/journal/summer04/patriot.cfm

[poisondeathray]

I agree with jagabo, the color suggests an individual layer issue .

themaster1's link explains about film layers , but reconstruction of an entire layer is extremely difficult - you will get color imbalances so that Patriot example is very impressive

Fortunately you don't have those issues, In your case, as mentioned earlier those blips, spots and scratches are quite easy to "fix" for "average joe" with free software (avisynth), at least compared to scratches in your earlier example which require more manual work or tracking . But combined with editing software (with separate tracks) and masks it can make your repairs better and more selective

...the most accurate/preserved output?

The most accurate output is a lossless format in the original colorspace (you are starting with a DVD, which is in YCbCr) , but most still image formats and compositing tools will be working in RGB . The earlier point was that when you convert YCbCr<=>RGB you lose quality, and if you're not careful in the conversion you might clip some data (in addition to the other losses)

[takearushfan]

I agree with jagabo, the color suggests an individual layer issue .

themaster1's link explains about film layers , but reconstruction of an entire layer is extremely difficult - you will get color imbalances so that Patriot example is very impressive

Fortunately you don't have those issues, In your case, as mentioned earlier those blips, spots and scratches are quite easy to "fix" for "average joe" with free software (avisynth), at least compared to scratches in your earlier example which require more manual work or tracking . But combined with editing software (with separate tracks) and masks it can make your repairs better and more selective

...the most accurate/preserved output?

The most accurate output is a lossless format in the original colorspace (you are starting with a DVD, which is in YCbCr) , but most still image formats and compositing tools will be working in RGB . The earlier point was that when you convert YCbCr<=>RGB you lose quality, and if you're not careful in the conversion you might clip some data (in addition to the other losses)

Thank you. I figured I probably was asking for far too much when requesting filter/script examples. I'm sorry but I still don't understand how Avisynth can be efficient enough to repair the kind of issues I've provided. I just said that I was requesting too much but now I'm basically asking along the same lines: What would an Avisynth script (even just the bare essentials) look like according to how it's possible for me to do these things myself?