Desray, I agree that nature and flesh tones are the best reference in terms of our opinions on accuracy. Flesh tones in particular are so subtly different, but our brains have had very good training in what is good and what is bad.
For Avatar, we may never truly have a reference, but it is still interesting to compare the different tones. I’m not sure if you read @Foodie message earlier, where I was asking him why his necklace was yellow and mine was red on our calibrated OLEDs. However, when we both watch the section with our eyes over and over again, it’s clearly a reddish orange and does not turn yellow at any point. The conclusion is simply the camera causing these color artifacts.
Since its impractical for us to do a Kansas City Shootout, perhaps we should run the same material in each others homes and watch together to see the differences as the best alternative. Each of us, Bryan, Fred, you and me have obsessively tuned our systems and are sure to be able to highlight what’s good and bad.
Agreed, choosing a capable camera, the CMOS, and the user’s ability to master the exposure triangle in photography will determine whether the image taken is true to the image quality presented to the one who took the image. There SHOULD NOT be any form of post-processing to enhance or otherwise make the image appear “different” from what one sees on the screen. I take it no one is doing that else the comparison will be pointless, to say the least.
There is no need to do that if all we are talking about is the difference in color rendition. What’s more important in my opinion is for one to believe there is an appreciable improvement in BEFORE and AFTER calibration. As long as your display has been calibrated using trusted calibration s/w tools like Calman, Chromapure etc even though it is an automated process for the most part. One can proudly say that his display has been properly calibrated.
There are certain calibrations that will not deviate. These are:
Black levels (optical and reference),
White levels using both Brightness (affects the shadow details in the lower IRE region) and;
Contrast (affects the upper IRE region where highlights can be clipped) to acheive it. White Balance and Gamma setting.
The next achievable settings are:
White Balance (RGB)
Doing a proper 1D LUT (grayscale to ensure there is no color fringing at either end of the IRE spectrum), minimally select 5x5x5 or best is 9x9x9
Gamma (select a suitable gamma setting before 3D LUT)
The hardest setting to do is color since not all display is capable of reproducing the colors within the BT2020 colorspace. Some may go all the way to calibrate for the full 9x9x9 which takes ages to complete and there is no guarantee that it will turn out well as the AE for colors may show you greater than 3.0 deviation error. Coupled this with different algorithms used by TV/Projector manufacturers which make color calibration neigh impossible. Not even the pro calibrator in the industry wants to delve into that “realm”.
I can only speak for the projector since I don’t do TV calibration. When I say Projector, my context here is referring to the conventional projector with proper housing and not those Ultra Short-throw (UST) projectors that claimed to have as high as 107% BT2020 colors! It’s almost laughable whenever I read about the marketing blurb. For the most part, JVC and Sony projector owners need not worry too much about 3D LUT since OOTB native color gamut can easily reach > 90% of the P3-DCI standards which in my books is pretty outstanding.
As you can see, I’m not touching on HDR tone-mapping as I do not consider this as a setting that can be achieved with a uniform standard due to the different algorithms used by different display manufacturers. Get the baseline BT 709 and BT 2020 in SDR right and leave the HDR tone-mapping to MadVR or Lumagen to do its proprietary thing - i.e. put it on auto-pilot mode and only engage manual override when you need more of highlights or shadow details, otherwise I will leave it intact for good.
Just to add on to my previous post on the differences in color tones for Avatar the way of water scene. I believe it all boils down in finding the “right spot” in the CIE chart and pin-point the HSL (Hue, Saturaution and Lightness) to ensure blue remains blue in its truest form and not over-exceeding to the point where blue becomes purple as show in this case.
Of course, the studio colorist plays a crucial role of how much to push the colorimetry.
See this 2017 seminar touching on HDR color grading if you have the time. Its very enlightening for those really into HDR color calibration.
Ya, that’s the thing. I’m really more interested in finding out which of those colours are accurate, true to the director’s intent. As close as possible is a good start. We are seeing so many different shades of green . I get that some would just argue, your device, your money, watch it however way you want it. Whilst that’s not wrong, personally for me, I’d prefer to follow a specific reference standard when it comes to video. For me, the colours must be accurate and true to the directors intent
I also don’t really like to boost the amount of luminance let through for an overly bright image, it can be really fatiguing when it comes to light sabre scenes. Some scenes are meant to be dark, if u r not supposed to see the shadow details, u r not… that’s the directors intent. Personally, I’d try to get my grayscale & gamma 0% -100% to track accurately, and that’s what matters most to me. But the strange part is all colours are accurately calibrated and yet we are still seeing variations in colour tone…
The issue on whether luminance is over boosted or not depends on the display device. Most movies are mastered to around 1,000 nits and some, all the way to 4,000 nits. For an OLED or LCD TV (some up to 2,000 nits), there is not much issue in displaying the full range of luminance. So when I watch on my OLED, I don’t need much tone mapping as the display device is very capable of displaying all luminance levels. All I need is some LLDV processing to tone map the 700-4000 nit content a bit and the rest looks fine.
The issue is with projectors that have, for example, max calibrated luminance of 50 nits. Heavy, heavy tone mapping has to be done to compress 0-4,000 nits into 0-50 nits. Some projectors like JVC and LG do a somewhat decent job. From my observation, when there are very bright parts in a frame like flames, both JVC Frame Adapt and LG DTM tend to make the dark parts too dark and you can’t see the shadow details. If you compare Frame Adapt or LG DTM to an OLED or high nit LCD display, the black crush is obvious. Hence, if you are looking for director’s intent, its about getting it to look as close as possible to the 4,000 nit Sony Mastering Display used by the editors. I think Lumagen and MadVR get closer to that than Frame Adapt or LG DTM. The net impact is that images with very bright and very dark parts are too dark on Frame Adapt and LG DTM, because they lower the average CLL too much to take care of the bright parkts, but look detailed and bright on Lumagen and MadVR because both systems employ more processing to dampen the highlights, while lifting the shadows.
Fore mere mortals like us, who don’t have a $100k Sony Mastering Monitor, the best reference is took compare projector output to a calibrated OLED or higher end LED display. The closer the look is to the OLED, the more accurate the tone mapping.
Here’s a youtube piece which looks at how MadVR deals with light sabre type scenes. Look at the section on Harry Potter Valdermoot battle MadVR tone mapping
My view is that the differences in color are due to cameras, white balance settings and projector contrast. The exception is where there were several images mixed in from SDR displays and I think the 20th Century Fox shots were rec 709, while most of the shots members posted were from BT2020 calibrations. SDR and 709 shots simply looked dull, while the BT2020 shots differed largely due to camera white balance and projectors’ contrast.
Finally above is the reference blue avatar from 20th century
The first thing which is obvious is the shade of blue, which all look different.
Then next, let’s observe the little boy”a face. On the first pic with madvr, we can see his face clearly. Second pic will be slightly darker and 3rd pic is the darkest
Whilst we can observe that the face is more visible with the madvr, we also note that the smoke in black in the background within the clouds, looks more grey on the madvr version.
If the boy is covered resting on avatar’s belly and not facing the sun (we can see his shadows on the avatar man), naturally that area will be dark. But from what I’m seeing, the face is equally as bright as the face of the avatar.
This is why I feel it looks less natural on the madvr version. Yes I get you want to see every detail on screen, but the question remains that some areas should just naturally be dark
It’s so much darker compared to madvr version below . But are you losing any details ? Not at all… you can see the shape of the vest of the man clearly…
I too was surprised that those Chinese projectors were claiming over 100% of BT2020 and had my doubts. However, delving in, many are using triple lasers and it seems that it is possible to get to more than BT2020 with the triple laser technology for short throws. The other thing is several USTs are now Dolby Vision certified. This is because they can reach the lumens and color gamut required by Dolby for certification whereas our long throw projectors cannot. Some long throw projectors have HDR10+ certification, because that is less strict than Dolby Labs.
Calman can now do Autocal for HDR calibration. What that means is that Calman’s Artificial Intelligence is able to calibrate HDR displays. However, there is a huge caveat in that its HDR colorimetry calibration (called Matrix LUT) uses just five reads, taking 30 seconds and it creates a 3D LUT for HDR using mathematics off these five reads. Portrait Displays says that this approximates the accuracy of a human manual calibrator . But just how accurate can a 30 second calibration be. In my experience doing HDR calibration for my Panasonic OLED, I achieved an average colorimetry DeltaE of 0.6, which is excellent, but the Panasonic was already well calibrated with a pre calibration DeltaE of 4.1 so Calman didn’t have to adjust much
Calibrating my LG HU810 projector was a different story. Calman’s Matrix LUT managed to get the average DeltaE down to just 2.3, but it started off at 20.6! The other thing is that most LG projector owners are unable to observe the quality of their HDR calibrations because LG’s internal pattern generator cannot do a post calibration measurement with Calman. I was able to do the following post calibration verification because I used an external pattern generator (PGenerator)
Bottom line is that when you calibrate HDR with Calman, you will not get an ultimately accurate color calibration (dE < 1.0) unless your device is already very good to begin with. LG projectors are not.
It’s a not a surprise to me as HDR base layer is derived from SDR 709/2020 depending on the color gamut used by the calibration software. This is why it is important and fortunately easy to get accurate BT 709 as base layer to start off as a basis point. The rest is merely expanding the colorimetry to fit within the bigger triangle in the CIE.
Agreed. I preferred the Century unadulterated SDR 2020 version over the MadVR but lest we forget, the source are different. We are using HDR-X version of the trailer which clearly have a different hues and shades as a result of the creative intent of the colorist from HDR-X. So it may not be an apple to apple comparison. Having say that, my Lumagen which converts HDR using PQ curve to SDR 2020 using gamma 2.4, it is quite natural and nowhere near as saturated as MadVR. This begs the question, is one source sharing identical HDR10 as base layer can have different color grading due to the difference in tone mapping and to a certain extent the camera used to take those images.
I’ll try to explain what is happening. MadVR dampens highlights and has an algorithm to make skies and clouds look more distinct while at the same time lifting shadows. This is done by photographers all the time in order to improve dynamic range. Humans see a much wider dynamic range than can be captured by cameras, so all sorts of processing is employed to improve the picture to what a human would see. This is supposed to be an improvement That’s not to say some people might not like it, but it is supposed to reflect a human ability.
LG’s DTM dampens highlights, but , unfortunately, applies it to the whole frame. Hence the dark parts simply get too dark. It makes no effort to lift shadows. If you get a chance look at this frame on an OLED or other high nit LED display, you will see that it looks more like the MadVR version than the LG DTM version.
I believe that the Century is not SDR 2020 but 4K Rec 709. This reflects the IMAX youtube trailer link that I posted. If you play it on your projector or TV, and hit info, it will say 4K Rec 709. That’s why it looks dull. HDR colors are just more vibrant
Maybe so if go by the technical aspect of things but one thing is for certain, unless some calibration software take it upon themselves to integrate say a Calman type plugin or software that can do proper calibration, the critical settings like black levels, white levels and colors will probably not be accurate due to the limited control settings to adjust it. It’s exciting news to see more manufacturers pushing the boundaries of technology to achieve what’s previously is impossible.
Exactly ! I agree, I personally prefer the Century image as well. They look very natural to me.
That’s why I intend to find out by comparing the original SDR version imax YouTube version… to have some insight on how different they are compared to the hdr version…
On that whale scene, the water looks more green than blue on century image… ours all look blue… hmm…
Anyway, it’s good to have an idea for comparison sake
Now the same photograph, but I used my iPhone editor to lift the shadows and improve the dynamic range, what MadVR tries to do. It now reflects more what I see with my eyes
. I get that some would just argue, your device, your money, watch it however way you want it. Whilst that’s not wrong, personally for me, I’d prefer to follow a specific reference standard when it comes to video. For me, the colours must be accurate and true to the directors intent
is with Madvr
is a reflection of the water
That’s not to say some people might not like it, but it is supposed to reflect a human ability.
