Sign In

DIF - Difference tool to spot visual changes - Current 2004 vs 2019

Author
Time
 (Edited)

Note: Starting to compare 2004 vs 2019 variations, jump to this post:
https://originaltrilogy.com/topic/DIF-Difference-tool-to-spot-visual-changes-Current-2004-vs-2019/id/13179/page/2#1340050

Previous DIF

Difference Blended Movie
aka Mechanical Approach to Spotting Visual Changes/Variations Between Film Versions

Currently : Star Wars Hexology 2004 vs 2011 Comparisons

Purpose : To identify visual changes made to the film, between these two versions.
Process : Two films version are composited using the ‘Difference Blending’ mode.  This renders pixels which are similar as black.  Areas which are dissimilar show up in color and stand out to varying levels.

Results: Comparison 2004 x 2011

http://www.archive.org/details/DIF-Star.Wars_2004x2011_480

WARNING Read full FAQ before watching.

FAQ Post: http://originaltrilogy.com/forum/topic.cfm/DIF-Difference-tool-to-spot-visual-changes-SW-2004-2011/post/538958/#TopicPost538958

Quick Pics Preview: http://www.archive.org/movies/thumbnails.php?identifier=DIF-Star.Wars_2004x2011_480

EDIT Previous First Post

Ok in my quest to make the most boring SW fan edit possible, I think i’ve stumbled across the idea. After seeing the side by side comparisons from the blu-rays of late:

(via: blu-ray thread)

This project will be named after the photoshop filter “Difference”. What this filter does is analyze two frame, where the information is the same it’ll return the color black, where there are differences it’ll return something else. So these edits will spot changes, and thus become more tools then edits. So i’ll use Tool from now on.

Example 1: R2 among the Rocks

Not the greatest example but you get a sense of about 50% of the frame being manipulated by the addition of the new rock.

Example 2: Ewok Eyes

This is where this Tool can pay off. It’s almost all black, the HD file is almost exactly the same as this supposed blu-ray leak file. The eye change being the noticable dissimilarity.

Here’s where I need help. No ones going to want to watch this whole thing, including myself. (I might, but doubtful) I need to find some video analysis program which has a feature which is a timeline and would show color peaks. Looking at this kind of chart it, should be able to identify some median line where no changes occur (ie the difference frame is mostly black) and the spikes would help us identify revisions.

Sound boring, yes it is!

“That’s the DIFFERENCE”

canned laughter

Author
Time

i would give this a watch too.  It would be nice so that you could tweek it a bit so that only new additions show up.  There are obvious slight color differences or minor shifts in the frame that will show up, but these really aren't changes.  You can see this on your preview of R2D2 - all the lines around him.  It would be nice if only the new boulders and shadow showed up in the print.  just a thought.

ThrowgnCpr’s edits on Fanedit.org

Author
Time

Sluggo wrote: I'd watch that once.

Tooooo kind, I wouldn't speak so loudly, someone might FORCE you to actually do it. heh heh heeeh.... *cough* hehehe BWAHHHHHAHAHHA Force you; get it?!

The final product should compress nicely.

ThrowgnCpr wrote: i would give this a watch too. It would be nice so that you could tweek it a bit so that only new additions show up. There are obvious slight color differences or minor shifts in the frame that will show up, but these really aren't changes. You can see this on your preview of R2D2 - all the lines around him. It would be nice if only the new boulders and shadow showed up in the print. just a thought.

All thoughts welcomed. New Additions Identification depends on the two sources used. Which could mean looking at looong stretches of blank black screen. (especially as i'm leaning toward HD and BR) Tweaking will happen, how it's applied is the 'art'. Which i'm not always the best judge of. Have a bunch of other screen caps to test with. For instance with the Ewok how do you pronounce the brown eyes out of the sea of black?

There's also a 'subtraction' blending mode.  The results lean to more blacks.  Here's a wikipedia explanation of the two:

Subtract

This blend mode simply subtracts pixel values of one layer with the other. In case of negative values, black is displayed.


Difference

Difference subtracts the top layer from the bottom layer or the other way round, to always get a positive value. Blending with black produces no change, as values for all colours are 0. (The RGB value for black is 0,0,0). Blending with white inverts the picture.

Trying a bit with the 2004s and a laser disc capture and I think this will be more benefitial and more successful results using the HD Broadcasts and the Blu-Ray. As on a pixel by pixel count they will be more alike for the difference to produce a true black.  The problem with these two sources is the HD broadcast might be the 16-236 (?) while the BR might be the full color range 0-255.

Also there's probably a different Difference which would leave white behind instead of black. The all black screen has it's appeal, but white might be easier to spot with.

Author
Time
 (Edited)

none said:

Tooooo kind, I wouldn't speak so loudly, someone might FORCE you to actually do it. heh heh heeeh.... *cough* hehehe BWAHHHHHAHAHHA Force you; get it?!

Is this one of your Earth jokes?

Author
Time

First Test Clip : http://noneinc.com/SWDIF/SWDIF_EarlyTests/SWDIF_480-Test_01.mp4 (31mb)

Ok with my machine only having 2GB of memory and me phucking up the order for more, i've stopped working with the full size frame and reduced to 480x204 to see if this whole project is worthwhile.  Seem to have figured out how to convert to the same frame rate, after a bunch of bad attempts.  This first clip focuses on the scene where R2 and 3po are just leaving the escape pod, people mentioned that the color of the pod lid has been revised.  And as you'll see in the clip, it stands out. 

See the blue dot.............................................................^  (exciting, I know!)  /sarcasm /BespinJokes

Another noticeable difference showing up is in the highlights off R2 and 3po:

.............................................^

This could be a drawback of the blu-ray source, but maybe they've upped some of the highlights toward a whiter white.  This is also visible at 00:14 in the clip a shot of one of the suns breaking over a sand dune.  Looks like the sky's been tweaked.

Now the caveat, is i'm learning the ropes here with attempting to maintain quality while trying to get a result.  So maybe when I transitioned from the source to my working 480 file, some things got shifted.  Will hopefully be able to compensate in future versions.

Right now taking the source files and running them through avidemux2.  Saving as motion-jpeg and avi format.  Then taking that and reducing to 480 in Quicktime.  Can eliminate that last step in the future.

Now i'm wondering if the black can be matted out and the 'difference noted' areas can be de-applied?

Author
Time

This was a great idea when it was first mentioned as a possiblity, and it still is. It also shows color changes, so we can PROVE that they haven't done any work on that.

Star Wars Revisited Wordpress / Facebook / Twitter

Star Wars Visual Comparisons WordPress / Twitter

Author
Time

As I recall, I cut the offending few frames (where R2's head and the scenery jolt - I assume it was when they got Kenny Baker out) and did a basic repair job using the MVTools plugin to create replacment frames.

DE

Author
Time

Is the top one the BD?

Also *cough* http://www.youtube.com/watch?v=Sce8inx_U2c

Yes, top is BD.  Seems like someone's taken a notice of your handy work.

To shorten the reviewing process, looking for something which will analyze the colors and making a simple chart of RGB over the video length, and frames which show an alteration will spike the chart, and then you can go in and watch that.  Otherwise i'm just running the movie and watching at 2 times speed, may miss something.

dark_jedi wrote: Can you share how you did the DE?

assuming your mean Darth Editous's fix not these DiffEdits.  If so:

http://originaltrilogy.com/forum/topic.cfm/The-Darth-Editous-Episode-IV-DVD-Info-and-Feedback-Thread-a-partially-de-specialed-DVD/topic/2425/

the youtube video says:

This new fix is now done entirely with about 10 lines of AviSynth script and one external plugin - the incredibly useful MVTools (version 2).

Author
Time
 (Edited)

To shorten the reviewing process, looking for something which will analyze the colors and making a simple chart of RGB over the video length, and frames which show an alteration will spike the chart


You could always use another filter to really boost the levels, then press shift-ctrl-[right arrow) in Virtualdub. In run through frames as fast as it can, and stop at what it thinks is a scene change (I think you can alter the threshold in preferences).

* though thinking about it, that wouldn't help if the change is brought in gradually.

DE

Author
Time
 (Edited)

That is alright, I do not need that link none, you can keep it, he answered right here.

Author
Time

So is this just to compare BD vs DVD?

It unfortunately wouldn't be possible to do with GOUT, as it doesn't match to the SE when overlayed (due to some sort of distortion) and the cropping changes from shot to shot (I know because I use the difference filter to match the two sources).

  • www.facebook.com/despecialized
  • IF YOU HAVE ANY QUESTIONS ABOUT THE DESPECIALIZED EDITIONS, PLEASE READ THE FIRST POSTS OF THESE THREADS, WHICH HAVE UP-TO-DATE INFORMATION: SW, ESB, ROTJ, 97SE RE-ED
    IF YOU DON’T FIND WHAT YOU’RE LOOKING FOR THERE, TRY ASKING IN THE APPROPRIATE THREADS - MOST REGULAR POSTERS KNOW ALL THE ANSWERS AND SOMEONE WILL LIKELY BE ABLE TO HELP YOU.
    IF I GET A PM WITH A QUESTION, WHICH COULD HAVE BEEN ANSWERED THROUGH THESE MEANS, IT WILL BE IGNORED. SORRY BUT I AM NOT THE LOCAL INFO BOOTH. THANK YOU.
Author
Time
 (Edited)

doubleofive wrote: So, are we saying this jump cut has been fixed for the Blu-ray?

Yes, R2 no longer rolls back before going forward.  As DE had fixed.

Harmy wrote:

So is this just to compare BD vs DVD?

It unfortunately wouldn't be possible to do with GOUT, as it doesn't match to the SE when overlayed (due to some sort of distortion) and the cropping changes from shot to shot (I know because I use the difference filter to match the two sources).

Right now i'm using the BD and HD broadcast sources, to test what I can accomplish, while also finding changes along the way.  The variations between GOUT and DVD are another ball of wax, as you point out.  They would require a two method approach.  Doing it once, straight no cropping, then a second pass to compensate for recropped scenes, compensate, then run again to see if other changes emerge at that point.  If there's warpage the results won't be as clean, but hopefully useful.

Having watched 30 minutes (beginning) and quick scanned another 30 minutes (the DS attack), there are few variations between these sources at the 1/4 quality level (480px).  It's not ideal, just as much as my sources are not ideal.  There's something weird going with the reds and bright white areas.  Like all the ceiling lights in the Tantive are registering.  Haven't run the full size test.

As Harmy says this is a common tool for aligning segments.  For me next step finding a program which can create the frame color analysis timeline is important as it will eliminate (possibly) the watching of this thing.  Mechanically finding changes can speed up the process in the future.

 

*EDIT*

the super shits'n'giggles DIFF release would contain multiple.  Theater Performance Recording vs CED or early VHS, later VHS vs LaserDisc, VHS vs GOUT, GOUT vs DVD, HD vs BR.  Each at full res, so it'd imply one of those logarithmic spirals:

yeah the resolutions don't equal diminishing values of the diagram... the later releases would be about changes, the earlier releases about quality improvements.  Never gonna happen, but might be nice to look at for a short spell.

Author
Time
 (Edited)

Learned the hardway that the German opening shot was updated differently for 2004/HD.  This file shows the difference between German HD and BR:

http://noneinc.com/SWDIF/SWDIF_EarlyTests/SWDIF_480-Test_02-FirstShot.mp4

A US HD broadcast cap is available here showing that this is not a specifc BR change: http://originaltrilogy.com/forum/topic.cfm/Complete-Comparison-of-Special-Edition-Visual-Changes/topic/11927/post/535009/#TopicPost535009

The different language crawls seem to be sliced into the film up to the first full screen white out explosion.

Author
Time

Can I just ask 1 simple question? what in the world are you trying to accomplish with this? I am having a hard time trying to figure out what all these analysis threads are actually going to do for you, I obviously must be missing something.

Author
Time

This one is great, actually, as it will show all the changes made to the Blu-Rays, which we may not notice otherwise (perfect for the kind of project I'm doing).

  • www.facebook.com/despecialized
  • IF YOU HAVE ANY QUESTIONS ABOUT THE DESPECIALIZED EDITIONS, PLEASE READ THE FIRST POSTS OF THESE THREADS, WHICH HAVE UP-TO-DATE INFORMATION: SW, ESB, ROTJ, 97SE RE-ED
    IF YOU DON’T FIND WHAT YOU’RE LOOKING FOR THERE, TRY ASKING IN THE APPROPRIATE THREADS - MOST REGULAR POSTERS KNOW ALL THE ANSWERS AND SOMEONE WILL LIKELY BE ABLE TO HELP YOU.
    IF I GET A PM WITH A QUESTION, WHICH COULD HAVE BEEN ANSWERED THROUGH THESE MEANS, IT WILL BE IGNORED. SORRY BUT I AM NOT THE LOCAL INFO BOOTH. THANK YOU.
Author
Time

dark_jedi wrote: what in the world are you trying to accomplish with this?

Initially, this is being done to spot changes in the recent set.  I don't have the visual knowledge of these films as many here do.  Never watched them as a kid as we didn't have a VCR or laserdisc.  So although i'm interested in the changes, I can't watch and spot, hey that's off.  But as everything is digital can approach finding changes from a mechanical perspective.  Let the computer figure it out.  But right now don't know how to get the computer to spit out the information.

Here's the graph i'd like to get: (no spit takes!)

Taking two sources and 'Difference' blending mode them, returns mostly black screen, so the chart would flat line most of the time, and the spots where a revision has occured will spike the chart.  Then just go to that spot in the video and see what's causing the anomaly.  That's the identifying part. 

 

The next part is where this could be helpful for people looking to de-evolve the films.  As we are hearing AotC has been recolored.  If this type of chart could be produced, my guess is the general settings for that color shift could be figured out and then unapplied to the film to take the BR and make it closer to the theatrical release.  If 75% of a sequence of scenes has a blue shift of 20% (+/- 5%), that can be dialed back.  The re-coloring projects have attempted this visually, this approach is numbers based.

If a full res DIF version is created, could be used as mattes to DeSpecialize things, or to create horrid video art.