- Time
- Post link
More like this ?
More like this ?
Yes, I noticed it too. The reality is, that any color balancing algorithm uses some criterium to balance the colors. Mine is no different. The second reel has a lot of red/orange in almost every shot, and almost no green, meaning that the colors are pretty biased, and therefore unbalanced:
It’s very difficult to find shots with sufficient amounts of red, green, and blue to calibrate the algorithm. I’m trying to figure out a way around this, but this is a bit of a challenge. The current algorithm is less sensitive to this problem, but as these shots show, apparently still too sensitive.
Am I the only one who thinks this image FEELS perfect? 😃
+1 😉
Yes, I noticed it too. The reality is, that any color balancing algorithm uses some criterium to balance the colors. Mine is no different. The second reel has a lot of red/orange in almost every shot, and almost no green, meaning that the colors are pretty biased, and therefore unbalanced:
It’s very difficult to find shots with sufficient amounts of red, green, and blue to calibrate the algorithm. I’m trying to figure out a way around this, but this is a bit of a challenge. The current algorithm is less sensitive to this problem, but as these shots show, apparently still too sensitive.
Am I the only one who thinks this image FEELS perfect? 😃
+1 😉
I don’t really think it’s “perfect”, 3PO is awefully orange in it. That said, because of you guys I made my own coloration of Star Wars blu ray with a high orange push and it looks a lot like this. Maybe I’ll post some screenshots of it in my thread on the freebie improvements I did on the Silverscreen edition.
That’s why I wrote it FEELS perfect and not LOOKS perfect 😃 What I meant was it feels like it was shot on a hot desert planet.
To me, the best pictures are the original attempt at the bottom half of DrDre’s first post. Those are perfect. I am not sure if all of these other ones came from different scans or whatever, but I like the original attempt the best.
This shows where DrDre’s tool really shines: as a base correction to later tweak.
Original:
obvious problems
DrDre’s:
much better, but too yellow
Corrected DrDre’s:
there we go
Mind you, that correction took me all of SECONDS to do. Whereas getting the original to that state would have taken considerably longer.
Original:
DrDre’s:
DrDre’s Corrected:
That’s why I wrote it FEELS perfect and not LOOKS perfect 😃 What I meant was it feels like it was shot on a hot desert planet.
GOTCHA… Your bases were fully covered by good grammar. 😃
To really run this “into the ground” though I did some experimentation.
I took Towne’s 2.7 and PUMPED up the saturation in VLC as well as a slight contrast boost:
Look at how close it is to the IB photo:
And simply for reference, here is Towne’s version with no image filtering:
What does that tell or teach us? In shows us how the RED/GREEN/BLUE levels can be spot on but there can still be tremendous differences visually because of how much Brightness, Contrast, and Saturation can effect the image.
I really could not understand why I’ve been getting such inconsistent results. I calibrated the latest version of the algorithm on this frame:
As far as I can tell, it looks just fine. When I apply the correction to other frames, I’ve been getting color shifts, that were mostly yellow/green, but also blue. However, when I examined the raw scan I noticed the relatively large differences in hue, brightness, and saturation between these two frames, that are only a few shots apart:
These inconsistencies are most likely caused by the scanning process, as poita’s scan looks much cleaner, an much more consistent.
That’s why I wrote it FEELS perfect and not LOOKS perfect 😃 What I meant was it feels like it was shot on a hot desert planet.
GOTCHA… Your bases were fully covered by good grammar. 😃
To really run this “into the ground” though I did some experimentation.
I took Towne’s 2.7 and PUMPED up the saturation in VLC as well as a slight contrast boost:
Look at how close it is to the IB photo:
And simply for reference, here is Towne’s version with no image filtering:
What does that tell or teach us? In shows us how the RED/GREEN/BLUE levels can be spot on but there can still be tremendous differences visually because of how much Brightness, Contrast, and Saturation can effect the image.
I think the v2.7 version looks perfect.
Here’s what my last corrected frame looks like with a boosted saturation:
What do you think about this tweaked shot ?
Looking good!
I’d say the sweet spot is about halfway between.
The same image minus the saturation adjustment :
Maybe what you could do is correct each shot to match the surrounding shots first and then use your tool to see if the results are more natural and consistent.
The same image minus the saturation adjustment :
Better! A bit “orange” to my eyes but if we’re going to tint SW, peach or orange is definitely better than blue, or magenta.
This is my take on that scene that I created by using avidemux to recolor the silver screen edition, using a “less contrasty” version of Dr. Dre’s correction as a visual base.
I made a thread on how I did it if you want to try it yourself… it’s become one of my favorite ways to watch the movie now when I’m not needing the clarity of Towne/Harmy’s 2.7.
http://originaltrilogy.com/topic/Improving-Silver-Screen-Edition-colors-on-the-cheap/id/49565
One of the things I use to judge the accuracy of any color correction is R2’s blue panels. Now, they don’t come out quite the same in every scene that was shot, but in the sources I trust, they are always blue, maybe leaning to a dark cobalt. So when I see a scene that shows R2’s with a greenish tint to him, I am convinced that it has been over corrected toward the green. One thing that tends to cause corrections to lean that direction is the skin tones can turn out too red for some tastes, but I’ve found that in the reliable reference photos, the skin tones when shooting in Tunisia were a bit red. When I correct the skin tones just based on the set photography, R2 usually stays blue and the people look a bit pink, but not the magenta of the Lowry HD work. This is my color correction of the UK GOUT:
Note the lack of green in the sand crawler and Owen’s hair as well as in R2’s blue panels. I find this color pallet to be my favorite.
One of the things I use to judge the accuracy of any color correction is R2’s blue panels. Now, they don’t come out quite the same in every scene that was shot, but in the sources I trust, they are always blue, maybe leaning to a dark cobalt. So when I see a scene that shows R2’s with a greenish tint to him, I am convinced that it has been over corrected toward the green. One thing that tends to cause corrections to lean that direction is the skin tones can turn out too red for some tastes, but I’ve found that in the reliable reference photos, the skin tones when shooting in Tunisia were a bit red. When I correct the skin tones just based on the set photography, R2 usually stays blue and the people look a bit pink, but not the magenta of the Lowry HD work. This is my color correction of the UK GOUT:
Note the lack of green in the sand crawler and Owen’s hair as well as in R2’s blue panels. I find this color pallet to be my favorite.
Well artoo’s panels do change quite a lot based on the different lighting, so merely looking to see if they are a blue color of some sort is not quite enough. But artoo is incredibly useful for judging color, particularly his dome and body. His dome we know is always a consistent aluminum silver color and his body consistently white, so all you need to do is zoom into his dome and body and look for color shifts/fringing that way. This does not account for the color grading of the scene, but getting the frame back to naturalistic colors is always a good place to start.
Well, in my experience, everything changes color depending on the lighting. On this case, it is sunlight, the identical light source found in a lot of on set photos. It also depends on camera filters and processing. But in the above frame I have barely touched the colors, mainly concentrating on the brightness, contrast, and saturation. It is pushed slightly to the yellow to reduce the reds, prevalent even in the GOUT.
One of the things I use to judge the accuracy of any color correction is R2’s blue panels. Now, they don’t come out quite the same in every scene that was shot, but in the sources I trust, they are always blue, maybe leaning to a dark cobalt. So when I see a scene that shows R2’s with a greenish tint to him, I am convinced that it has been over corrected toward the green. One thing that tends to cause corrections to lean that direction is the skin tones can turn out too red for some tastes, but I’ve found that in the reliable reference photos, the skin tones when shooting in Tunisia were a bit red. When I correct the skin tones just based on the set photography, R2 usually stays blue and the people look a bit pink, but not the magenta of the Lowry HD work. This is my color correction of the UK GOUT:
Note the lack of green in the sand crawler and Owen’s hair as well as in R2’s blue panels. I find this color pallet to be my favorite.
I definetly like these colors a lot 😃
Well, in my experience, everything changes color depending on the lighting. On this case, it is sunlight, the identical light source found in a lot of on set photos. It also depends on camera filters and processing. But in the above frame I have barely touched the colors, mainly concentrating on the brightness, contrast, and saturation. It is pushed slightly to the yellow to reduce the reds, prevalent even in the GOUT.
Well yes, but my point was that while lighting does change everything, under no lighting (especially not direct sunlight) should there be a, say, green tint to artoo’s dome (unless lit by a green light, which is incredibly unlikely in this particular film).
A test of a global lightening of the sample (it is not a proper gamma correction). Dark shots are still very dark (interesting to compare with Harmy’s choice here : https://lh3.googleusercontent.com/-P9zKZRT5Kvw/UlNIHMlr-UI/AAAAAAAACbw/nGWESqsmlLIDnElbq395IC9DwEfkwtNxQCL0B/w997-h561-no/Compar%2Bmaster.avi_snapshot_03.26_%5B2013.08.01_11.12.45%5D.jpg).
One of the weaknesses of the color restoration/balancing algorithm, is that it requires sufficient amounts of red, green, and blue (or their mixes) to be in a frame or image, for it to give accurate results. The Luke meets Obi-Wan frames are particulary challenging, because the colors are supposed to be skewed towards the oranges. I’ve been considering how to deal with such frames or sequences. One way to deal with this, is to find areas in the frame that are supposed to be balanced, namely the grey/white areas. Calibrating the color restoration/balancing algorithm on the white/gray parts of R2-D2 for example gives much improved results:
Deleted