Quantcast Color Correcting and Removing Artifacts from 8mm Film Scan - digitalFAQ Forum
Go Back    Forum > Digital Video > Video Project Help > Restore, Filter, Improve Quality

Reply
 
LinkBack Thread Tools
  #1  
06-01-2018, 04:00 PM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
I've had some time to get busy recently and I've scanned all 30 rolls of 8mm movie film which my Mom's good friend gave me to practice on. I'd like to do more than to hand her (and future customers) the scanned raw footage. At the same time, though, unless someone is willing to pay a premium for extra attention I can't afford to spend too much time on any one reel or scene.

The source material for the attached files is consumer-grade 8mm home movies...no special attention to shooting or storage. They were captured uncompressed with a MovieStuff RetroScan Universal film scanner; native resolution 1296 x 964 pixels. After capture they were exported as uncompressed 1080p AVI files; I have selected the attached clips using VirtualDub and saved lossless as HuffyUV but have not performed any post-processing or color correction as of yet.

The RetroScan software gives me the choice of exporting to 1080p or to 720p; nothing in-between. I chose to upscale to 1080 with the thought that I would make some efforts to correct color and remove artifacts, crop the frame slightly, add some royalty-free background music with VideoStudio and then render the final product at 720p for delivery to Mom's friend as digital or Blu-Ray or else downscale it to 480p if she prefers a DVD.

I don't have a problem with the editing and such, but I don't have an artist's eye for color correction and I'm a newbie at using filters in VDub. I did pay for VDub's Neat Video plugin last year so I have that available; I know our resident Smurf is not fond of it but I wouldn't mind getting some mileage out of the purchase. I'm also completely inexperienced at using histograms and such to optimize color correction. So I thought I would throw these out to see what recommendations our board experts could make, as well as pointers to any tutorials I might have missed.

Clip 1 is their family in a field of bluebonnets. The colors have faded and there are artifacts from the film stock degradation, but resolution is really not too bad.

Clip 2 is from Christmas 22 years ago. Obviously filmed under low light conditions; a very poor original image.

Clip 3 is actually the oldest of the lot, from about 1972. It was filmed on Regular 8mm as opposed to the Super 8 of Clips 1 & 2, but it really does have the best image of the three. I thought I would include it to show what the RetroScan is capable of when it has good raw material.

Thanks in advance for any education and pointers.


Reply With Quote
Someday, 12:01 PM
admin's Avatar
Ads / Sponsors
 
Join Date: ∞
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
  #2  
06-01-2018, 07:51 PM
FredGarv FredGarv is offline
Invalid Email / Banned / Spammer
 
Join Date: Nov 2017
Posts: 1
Thanked 0 Times in 0 Posts
I'm impressed by the 3rd scan. The 1st could use some chroma boost, but it looks like the green from the foliage is all but gone. Might be tough to restore.

The 2nd could do with neat nr. Some say it reduces rez, but my scope says no, if you don't overdo it. Let your eyes be the judge of that. Artificially adding a bit of sharpness after neat has helped me. I like to leave a little noise/grain in my film scans.

Not sure what your software can do, but I think these can be adjusted to look quite nice.

I had some super8 scanned at 2k a couple of years ago. It was OK, but there were some issues. Perhaps I should have spent the money on one of moviestuff's scanners. Will have to check out his offerings, haven't been to his site in quite a while.

Good luck, let us know how they turn out.

Last edited by admin; 06-02-2018 at 06:47 AM. Reason: User has invalid email, must Contact Us to resolve.
Reply With Quote
  #3  
06-02-2018, 01:12 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
Here's what I managed to accomplish playing with the first reel, the bluebonnets. I used two filters in VirtualDub; first the ColorMill plugin to give it just a little bit of a boost in the green and raise the saturation and brightness level a hair. To be honest I wasn't familiar with VDub and I had been working on the reel scene by scene but when I saved it I was using the settings for the very last scene when I was trying to bring out the grandmother's face from the backlight; I lost the settings I had come up with for the earlier scenes. However, after it rendered it came out pretty darn good; I didn't go back and redo the opening scenes. Then I used Neat Video for noise reduction and cropped to 1392 x 1040.

I took the AVI out of VDub and went into VideoStudio. There I edited, inserted the opening clip, and added background music using VideoStudio's royalty-free "Auto Music" tool. Then I rendered to mp4 at 720p and 480p. The 480p clip fits in under the 99MB limit in its entirety; I'm also attaching a preview outtake of the 720p clip so that you can compare quality. Also for comparison I'm attaching a 480p render of the raw capture, no processing, so that you can see what I had to work with originally.

Again, all of the capture and intermediate processing was performed lossless right up to the final render, so I can go back and make changes if anyone has helpful suggestions. Let me know what you think!

-- merged --

For the "Little Girl In Pink" reel, which was good from the original scan, I just used the HSV filter plugin to make some very minor tweaks, then cropped to 1392 x 1040 and used Neat Video to remove noise. Again, I added titles and background music in VideoStudio and rendered to mp4 (and also mp2 for later inclusion on a Blu-ray). I'm attaching the full reel in 480p, a sample clip of the 720p, and again a 480p render of the raw capture.


Attached Files
File Type: mp4 SG029-LittleGirlInPink.480.mp4 (78.27 MB, 23 downloads)
File Type: mp4 SG029-LittleGirlInPink.720clip.mp4 (83.61 MB, 12 downloads)
File Type: mp4 SG029.raw.480.mp4 (46.63 MB, 9 downloads)
Reply With Quote
  #4  
06-02-2018, 02:23 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
This one I'm proud of! I originally thought that I'd have trouble getting the Christmas reel to look good. And, indeed, whatever I did in VirtualDub to correct the color ended up with it looking worse than the original. I ended up simply reducing noise with Neat Video and exporting it as is.

However, in VideoStudio I read the freaking manual and learned how to use "white balance" to color correct a clip. I picked out a white area...the piece of paper the little girl was holding served nicely...and then I played with the color temperature settings. Tungsten (2800K) was just a little too warm, while fluorescent (3800) was just about right. I was able to drop it a bit, to 3200K, and that's what you see in the attached clips. I think they came out pretty good, if I do say so myself.

Original raw capture file rendered to 480p.

Full processed reel rendered to 480p.


Clip of processed final version at 720p.


Attached Files
File Type: mp4 SG016.raw.480.mp4 (65.19 MB, 26 downloads)
File Type: mp4 SG016-Christmas.480.mp4 (58.86 MB, 16 downloads)
File Type: mp4 SG016-Christmas.720clip.mp4 (69.47 MB, 9 downloads)
Reply With Quote
  #5  
06-03-2018, 02:48 AM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
Quote:
Originally Posted by ehbowen View Post
The RetroScan software gives me the choice of exporting to 1080p or to 720p; nothing in-between. I chose to upscale to 1080 with the thought that I would make some efforts to correct color and remove artifacts, crop the frame slightly, add some royalty-free background music with VideoStudio and then render the final product at 720p for delivery to Mom's friend as digital or Blu-Ray or else downscale it to 480p if she prefers a DVD.
Kudos for all the work you're apparently doing. I get tired just thinking about it. I'd think 720p might be a better choice, since it's easier to upscale that to 1080 for Bluray, and much better to downscale to 720x480 for DVD than it is to downscale 1080p to DVD frame size.

You will encounter some problems: 1080p at 17fps is not valid for BluRay, and neither is 720p at that frame rate. You can approach DVD by encoding elementary streams at 19.98 fps and applying pulldown to bring the frame rate to 29.97fps. That works for DVD using the DGPulldown app, but I think you might need special software (expensive?) to get BluRay up to telecined 29.97 from a 17fps original.

It seemed to me that when I downscaled and worked with your Sharon Green sample, the resulting telecined 480i video (attached) looked sharper on playback than did the big-frame HD original. But maybe that particular video just "looks that way" and the others won't. You never can tell. The resizer was an Avisynth routine especially designed to avoid many downsampling distortions. I used an Avisynth contrast filter to bring out mom's shadowed face a little. I don't think you'll get believable full color from the greyed-out image. But you could salvage it with a little denoising to maintain and enhance what little detail is in the capture (I used Avisyntyh) and make it a clean black-and-white record, which I think looks better than corrupt color and isn't as distracting. But you might prefer another solution.

The attached m2v is encoded for DVD and does indeed run at 29.97fps with pulldown applied, but the base film had to be encoded at 19.98 fps instead of 17 and the resulting m2v is effectively interlaced video. I also applied some stablization using VirtualDub's DeShaker filter, but it does mean losing a few pixels along the borders. Keep in mind that HD BluRay at 29.97fps and standard definition BluRay at 29.97fps must be encoded as interlaced.

Quote:
Originally Posted by ehbowen View Post
I don't have a problem with the editing and such, but I don't have an artist's eye for color correction and I'm a newbie at using filters in VDub.
I think you did well, especially for not having an eye as you put it. I don't think anybody's born with that eye, at least not so ready-made that they're able to use it without training and practice. Your posted results deserve a lot of credit.


Attached Files
File Type: m2v Sharon_HC2_bw_DVD_Pulldown.m2v (848.2 KB, 8 downloads)
Reply With Quote
The following users thank sanlyn for this useful post: ehbowen (06-03-2018)
  #6  
06-03-2018, 06:08 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
It looks as though I'll need a course of sprouts in frame rates and encoding. Should that be its own thread?

By the way, the 17 fps frame rate is the RetroScan's default export setting; it can be tweaked up or down as needed. It actually saves each frame as an individual still picture and you can export as a series of .jpg or .tif images instead of a video stream if desired. My understanding of 8mm film is that silent Regular 8mm runs 16 fps in source and silent Super 8 runs 18 fps, so I left the setting at the default.
Reply With Quote
  #7  
06-03-2018, 08:14 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
Additional Info: Here's a screenshot of the software export page showing the different frame rate options. They don't affect the actual scanned content, of course, only the rate at which the file is instructed to display. Can you recommend one of these options as preferable for my purposes (possible future export to DVD/Blu-ray)?



Attached Images
File Type: jpg Capture.20180603.jpg (60.4 KB, 126 downloads)
Reply With Quote
  #8  
06-03-2018, 09:42 AM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
I see two choices as reasonable:

1440x1080 (4:3) @23.976 is valid for high definition BluRay. It will be progressive video, which is also valid for that frame size and frame rate for BluRay/AVCHD. This choice could be downsampled as progressive video for 720x480 DVD or standard def BluRay as 23.976 film speed, and most encoders can add common 3:2 pulldown for 29.97 standard definition NTSC DVD/BluRay/AVCHD.

Note there is no 4:3 ratio allowed for HD BluRay. 4:3 HD video is displayed as 16:9 with side pillars around 4:3 video. 4:3 is allowed for 720x480 standard definition Bluray.

As 1440x1080 23.976 progressive video it can be downsampled for internet posting although some websites might balk at the frame rate (they like either 25 or 30 fps, and you can't post telecined or interlaced video on the net). If you chose 1440x1080 @29.97 fps (I don't know how your capture device does that) it would be progressive video that is not valid for BluRay at that frame rate and frame size -- but you could encode it with fake interlace flags to be HD BluRay compliant. 1440x1080 could be downsampled directly to 720x480 @29.97 and encoded with interlace flags for DVD or SD BluRay. It could also be downsampled to 640x480 29.97fps progressive for internet posting.

The resizers I used are Avisynth algorithms and filters, notably Spline36resize with chroma shift parameters for the type of video and downsampling. NLE's, including "Pro" types, use much softer or much sharper (with visible artifacts() resizers.

720p formats for HD BluyRay/AVCHD are strictly progressive at either double frame rate or film speeds. Pulldown and interlace are not allowed. 16:9 only.

Basic BluRay spec (NTSC and PAL): https://www.videohelp.com/hd#tech
Basic DVD spec (PAL and NTSC): https://www.videohelp.com/dvd#tech
Encoding for h264/AVC/MPEG BluRay formats (good encoders enforce these specs): https://forum.doom9.org/showthread.php?t=154533

Any other ideas out there?

Last edited by sanlyn; 06-03-2018 at 10:13 AM.
Reply With Quote
  #9  
06-03-2018, 10:11 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
The thing is, playing (original) 16-18 fps video at 23.976 fps is going to look like an outtake from the Keystone Kops. What I need is a frame rate which can be converted to a legal frame rate in a way which doesn't overly distort the original 's quality of motion. Any suggestions?

By the way, the Retroscan software can export to 16:9 aspect ratios as well, either by cropping or by padding/ letterboxing. Should I try an example or two for practice?
Reply With Quote
  #10  
06-03-2018, 10:28 AM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
Quote:
Originally Posted by ehbowen View Post
The thing is, playing (original) 16-18 fps video at 23.976 fps is going to look like an outtake from the Keystone Kops.
I see. I was getting the idea that your capture device simply inserted extra frame to get different frame rates. Yes, speedup definitely would look ridiculous. The m2v that I posted earlier was encoded at 19.98 fps because the software I'm using is limited to adding pulldown at frame rates no slower than that, so the result was a slight speedup from the original 17fps you stated. If your silent films are running at 18fps, a speedup to 19.98 wouldn't be all that noticeable. If you had sound film at 16fps, a speedup from 16 to 19.98 is going to be obvious to most viewers.

I'm certain there is software out there that can properly add the necessary fixes, dupe frames, pulldown, etc., that can get you into standard encoding territory but it's not anything I've used myself. Obviously it exists -- look at all the 16 to 20 fps classic silent films that have been encoded to DVD and Bluray.

There are bound to be readers out there with more expertise in this area.
Reply With Quote
  #11  
06-03-2018, 10:43 AM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
The Retroscan captures at 10 fps on slow speed, or 15 fps at high speed. This is a function of bandwidth limitations, especially of the USB2 connection from the camera to the capture computer. (A 2k HD camera has recently become available as an upgrade, but it requires USB3 and a high-speed RAID array or SSD.) The software doesn't know what to export at until you tell it.
Reply With Quote
  #12  
06-03-2018, 07:34 PM
ehbowen ehbowen is offline
Free Member
 
Join Date: Feb 2018
Location: Houston, Texas
Posts: 341
Thanked 64 Times in 57 Posts
Why am I worrying? VideoStudio automatically converts to a legal frame rate when I use it to render! Check the details of the finished files I included above...29.97 fps and motion looks quite normal. Now, there is some perceptible jerkiness which is probably due to the oddball 17 fps frame rate I started with, so I might want to see if 16fps or 18fps converts to a smoother end product. But I don't need to lose any sleep over this question.
Reply With Quote
  #13  
06-03-2018, 10:35 PM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
I'm not familiar with your capture device, but it's a good break that it works for you.

There are some serious but manageable color and signal level; problems in the pink girl and christmas videos. You managed to overcome several problems quite well, but I'll prepare a demo of suggested corrections that you can use for these and other videos. Will return later in the A.M.
Reply With Quote
  #14  
06-04-2018, 03:31 AM
spanak spanak is offline
Free Member
 
Join Date: Sep 2017
Location: Blugaria
Posts: 23
Thanked 4 Times in 3 Posts
Quote:
Originally Posted by ehbowen View Post
Why am I worrying? VideoStudio automatically converts to a legal frame rate when I use it to render! Check the details of the finished files I included above...29.97 fps and motion looks quite normal. Now, there is some perceptible jerkiness which is probably due to the oddball 17 fps frame rate I started with, so I might want to see if 16fps or 18fps converts to a smoother end product. But I don't need to lose any sleep over this question.
16fps or 18fps is still not enough for smooth video.

If you really want to smooth it out, you need to interpolate to a higher frame rate. Modern TV sets are able to do that, however I am not sure if the can handle frame rates that low. The other option I can suggest is to use ffmpeg, like this:
Code:
ffmpeg -i 8mm-scan.avi -filter "minterpolate='fps=29.97'" -vcodec huffyuv interpolated-output.avi
Or, if you prefer not to interpolate, but only convert frame rate and keep video speed:
Code:
ffmpeg -i 8mm-scan.avi -r 29.97 -vcodec huffyuv output.avi
Of course, you can use any frame rate, codec and output container you like.

I am attaching a sample of video 029 interpolated to 29.97 fps, and a sample that has only been converted to 29.97. Judge for yourself what is better.


Attached Files
File Type: avi 029-29.97-plain.avi (83.18 MB, 5 downloads)
File Type: avi 029-29.97-interpolated.avi (69.32 MB, 13 downloads)
Reply With Quote
The following users thank spanak for this useful post: sanlyn (06-04-2018)
  #15  
06-04-2018, 05:52 PM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
Thanks for the work you must have put into mounting all these samples. It's been instructive for me as well.

Quote:
Originally Posted by ehbowen View Post
I'm also completely inexperienced at using histograms and such to optimize color correction. So I thought I would throw these out to see what recommendations our board experts could make, as well as pointers to any tutorials I might have missed.
What the experts would recommend is learning to read histograms. Working without them is like working in total darkness. Below are two free links. They deal with still images, but the graphics principles apply to photos and video alike. After all, video is a stream of still images:
Understanding histograms Part 1 and Part 2
http://www.cambridgeincolour.com/tut...istograms1.htm
http://www.cambridgeincolour.com/tut...istograms2.htm

Your 29.97 fps samples apparently already have some framing adjustments applied. I'm surprised you haven't viewed your videos frame by frame. They contain a parade of duplicate and near-duplicate frames inserted to adjust the frame rate.

At first I wasn't going to try a color correction demo because the only sizeable sample that seemed close to the original capture's color was apparently "SG016.raw.480.mp4". But it's misnamed as "raw," which it isn't -- it's a rather bad low-bitrate and very lossy encode of the original with a lot of compression artifacts, bursts of mosquito noise with motion, and distorted scene transitions (did you use your NLE to make cuts after encoding the mp4?).

But it occurred to me that this would make a useful demo to illustrate why lossy encoding shouldn't be used for restoration work. I did some denoising but not enough to scrub all the compression noise, which would be futile. So in the attached rework from the "raw' mp4 you'll see noise remnants. The noise being what it is, It won't go away without eviscerating the video itself. I realize that you used NeatVidedo to curb what you thought was grain, but that noise isn't grain. It's CMOS residual processor noise. What it represents is tiny areas where no video data exists. I wish I had a travel refund for all the mileage I've put into my own NeatVideo for years since version 1.1. But in this case CMOS noise is not what you want to us it for. The Christmas video looks blurry and plastic, something that's easy enough to avoid with NeatVideo but not without help from other filters.

Be that as it may, maybe the notesband picrtyures below will help with all your videos, not just the Christmas project.

The first step I took was to work in YUV to make basic corrections to signal levels and color balance. This meant working in the sample's base YUV, in this case YV12. YUV is a video storage system that stores luminance information in the Y channel, blue-yellow in the U channel, and red-green in the V channel. You can modify channels separately. The YUV value range range for most flavors of digital video is stated as y=16 to 235, where 16 is the lowest or darkest value and 235 is the brightest or highest value. I know you've heard of the range 0-255 for color, but you're thinking of RGB. RGB is display technology; YUV is storage tech. You'll have to go back to the earliest days of TV to recount where YUV and 16-235 came from. Suffice it to say that YUV is still in use.

The trick to remember is that YUV's 16-235 value range is expanded to 0-255 when YUV is decoded and played as RGB. RGB expands the darks beyond 16 and expands the brights beyond 235. Now, what happens if your YUV video already exceeds 16-235 (which, in fact, YUV can do). For example, what if it's already 0-255 YUV to begin with? Simple. Out-of-range range values are clipped off (destroyed). Once clipped, those details are gone forever.

There's not much clipping in the Christams video, except of course with the tabel lamps and such. There's a lot of it in the little girl video. Film can accept a far greater contrast and color range than digital video. How you could tell your scanner to handle that, I can't know.

The image below is an unfiltered but slightly resized frame 2393 from the original "SG016.raw.480.mp4". To the right of the image is an Avisynth YUV levels histogram schematic of that frame:



What does that schematic tell you? The top horizontal band (white) is the luminance or Y-channel band. Darkest values are to the left, brights are to the right. Note that there's a shaded left and right border area on each side of the graph: the shaded borders represent unsafe values below y=16 or above y=235. The middle colored band is the U channel (blues and yellows), the bottom band is the V channel (red/green). Folks like to call the U channel Blue and the V channel Red. I don't know why. Maybe it's because Blue and Red are the only values stored directly in YUV. Green is derived by subtracting U and V from Y. Saves data storage space.

The YUV histogram clearly illustrates that there are a whole lot of red pixels and little or nothing from yellow and green. Other color data is there, all right, but it seems everything is lodged to the right of the middle mark. So you've got some really distorted color. In this and other shots the white Y levels are creeping into the unsafe zone. A little right-hand "peak" indicates some moderate clipping in the source itself.

Below, this is what happens when I use Avisynth functions to shift all color values toward the right on the YUV color scale. This affects color only, not luminance, so I used other commands to compensate for the perceived lowered brightness by boosting Y contrast a very small amount.



This correction at least gives a more workable color range. As the image above shows, colors look more natural but darks are a bit greenish and a few other areas can be tweaked later in RGB with VirtualDub. RGB filters in VDUb and in bigger NLE's like Adobe Pro or Vegas Pro, etc., allow sophisticated adjustments such as modifying very specific hue and brightnmess ranges without affecting other parts of the image. Such precision is extremely difficult in YUV without very complex (and expensive) programmed controls. Besides, working in RGB is more intuitive IMO.

The VirtuaLDub filters I used for mild color tweaks were gradation curves and ColorMill, which are similar to curves and color wheels in pricey NLE's. Below, here arev a few frames comparing the color in your "SG016-Christmas.480.mp4" (left image) with the colors in the attached Reworked mp4 (right image). The frames have been reconfigured as 4:3 images to show how the image content displays when played:

Frame 2363 (left) from "SG016-Christmas.480.mp4" -vs- frame 2093 (right) from "mp4_Reworked.mp4":


Frame 2768 (left) from "SG016-Christmas.480.mp4" -vs- frame 2492 (right) from "mp4_Reworked.mp4":


From the SG016 images it's easy to see that the original red color cast has been replaced by a purplish-blue one. This is a common technique with newcomers, but it's the wrong solution. Red's opposite balancing color isn't blue, it's cyan (blue + green, a secondary color). Adding blue to red simply makers it purple. Primary colors never have another primary as their opposite.

There are hundreds of ways to balance colors in that video. The trick is to decide on what colors certain objects should look like. Obviously the tree should have some green in it. The piano in the background of the living room sofa scenes is probably a darkish brown (you can barely see its outline in the shadows, but we know the piano isn't blue or green). It helps to know that brown is really a reddish red+green mix darkened by some dark blue. You can test piel colors in objects using a pixel reader. One such desktop freebie for Windows is ColorPic (http://www.iconico.com/colorpic/).

Better denoising in the rework would be highly desirable, but that's not possible with the lossy sample. In the attached Rework mp4 I discarded dozens of broken and distorted transition frames. At some points compression distortion was so bad that faces were obscured and objects on walls would disappear and reappear. The mp4 is slightly bright to demonstrate the kind of dark-area floating grunge you see in low-bitrate encodes. The murky noise actually change color from moment to moment.

I think the final overall result has just enough leftover warmish pale yellow to transmit the warm balance of interior lighting. One could probably tweak the exact colors and levels forever. The intention is to make scenes believable so that the viewer gets past defects and the scenes become less boring, less annoying, and more interesting, and the people take on more life-like attributes.


Attached Images
File Type: jpg 2393 YUV original R.jpg (85.5 KB, 109 downloads)
File Type: jpg 2393 YUV after R.jpg (84.9 KB, 109 downloads)
File Type: jpg 2363 480 mp4 vs 2093 reworked mp4.jpg (64.4 KB, 108 downloads)
File Type: jpg 2768 480 mp4 vs 2492 Reworked mp4.jpg (80.7 KB, 107 downloads)
Attached Files
File Type: mp4 mp4_ReWorked.mp4 (91.24 MB, 1 downloads)
Reply With Quote
  #16  
06-04-2018, 09:01 PM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
OOPS. I forgot...

To reply to a specific part of this thread's title, "Removing Artifacts from 8mm Film Scan ":
Artifacts are digital creatures. Film is analog. Analog doesn't have digital artifacts. Analog can have all kinds of other problems, but not digital artifacts. If your scans or work files have artifacts, they are digitally created and have to be added to the original signal by digital processing somewhere along the line.
Reply With Quote
  #17  
06-05-2018, 02:32 AM
spanak spanak is offline
Free Member
 
Join Date: Sep 2017
Location: Blugaria
Posts: 23
Thanked 4 Times in 3 Posts
Quote:
Originally Posted by sanlyn View Post
What does that schematic tell you? The top horizontal band (white) is the luminance or Y-channel band. Darkest values are to the left, brights are to the right. Note that there's a shaded left and right border area on each side of the graph: the shaded borders represent unsafe values below y=16 or above y=235. The middle colored band is the U channel (blues and yellows), the bottom band is the V channel (red/green). Folks like to call the U channel Blue and the V channel Red. I don't know why. Maybe it's because Blue and Red are the only values stored directly in YUV. Green is derived by subtracting U and V from Y. Saves data storage space.
The reason to call color channels U and V is quite simple. As you can see YUV picture is a 3-dimensional space, because each pixel has three coordinates - Luminosity, Red-Green and Blue-Yellow. Of course, this is from visual perspective, we are not talking about pixel position, nor timing. An entire video clip would be a series of vectors in 6D space: (Luma , Chroma1, Chroma2, Position1, Position2, Time).

Traditionally, in geometry (x, y, z) are used for coordinate names. But when referring to one of the planes as a part of the 3D or (more dimensional) space, usually U and V are used, to avoid ambiguity. Hence we get a UV plane.

Same applies to YUV space. Colors form a plane, which is part of the entire space, so we use U and V coordinates (color values), instead of X and Y. In fact, calling it an XY plane would have caused even more ambiguity. What are X and Y? Are we still talking about colors or position on the screen?

I hope I explained it clear enough. If not, refer to stackexchange.
Reply With Quote
  #18  
06-05-2018, 03:49 AM
sanlyn sanlyn is offline
Premium Member
 
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,266 Times in 973 Posts
Quote:
Originally Posted by spanak View Post
I am attaching a sample of video 029 interpolated to 29.97 fps, and a sample that has only been converted to 29.97. Judge for yourself what is better.
Thanks for the samples. Actually tougher vto decide than I thought.

It appears that the "raw" Christmas mp4 from post #4 has duplicate and near-duplicate frames to make it run at 29.97fps.
Reply With Quote
Reply




Tags
8mm film to dvd, cleanup, color correction

Similar Threads
Thread Thread Starter Forum Replies Last Post
Removing horizontal/vertical film lines and scratches? yukukuhi Restore, Filter, Improve Quality 139 06-17-2018 08:15 AM
Color band artifacts during capture benzio Capture, Record, Transfer 3 04-04-2018 12:06 PM
Help correcting de-interlacing problems, please. TerryJW Restore, Filter, Improve Quality 1 06-21-2016 04:25 PM
Advice for a film scanner for 35mm B&W negatives and color slides? Gabriele83 Photo Processing, Scanning & Printing 3 01-31-2015 06:52 AM

Thread Tools



 
All times are GMT -5. The time now is 06:05 PM