[Daniel.N]

Assume that we agree that for digital connections between source and display device..


a) video levels should be maintained

b) video is intended to end up within the 16-235 interval

c) stuff that ends up below 16 should NOT be visible.

d) stuff that end up above 235 SHOULD be visible

e) to keep this discussion cleaner, lets pretend we have perfect D65 tracking whereever we set the contrast


For the sake of THIS discussion, please refrain from discussing points a to c or e. Point d) is the one I'm after here. Fair enough, I've paused my player in lots of scenes where I'd expect to find points of maximum white and sure enough, a contrast setting so high that values above 235 can't be distinguished make these scenes look bad. In my experience, resolution above 235 is clearly needed, and this so often that setting maximum white of my display to match 235, although giving a "punchy" nice image, will be easy to spot since clouds start to look like marshmallows etc. So, point d) is as clear as a) to c) I think.


Getting to the real point then.. How MUCH above 235 _needs_ to be distinguishable? There is data above 235 quite often, but is there EVER any data above, say, 245? 250? 240? I'm not after the "should" here, I'm after the reality-checked value to shoot for.


I know that you "should" be able too distinguish values all the way up to 254, but maybe you'll end up reserving the maximum white of your display for values that are only present in that one scene in that one film that you don't watch very often?


I can live with better contrast in 99% of my movies even if the 1% look worse. If it's 80-20, I could not. Anyone got hard numbers or experience to share here? Thanks.

 

[ChrisWiggles]

What is your display?


A theoretically ideal accurate situtation would maintain everything to 254. I personally do not particularly advocate this per se. I suggest you view content and bring that point down progressively and see how it affects things empirically. And arrive at your own compromise point based on your testing and preferences. If you have a digital with a limited on/off CR range raising the white level (lowering the white point in terms of the digital range) will increase the CR of scenes, but you may loose a little bit of stuff above white, if it is present. If it's not present in the content, you're not losing any whites but you're losing CR. It's a tradeoff, and while I do not at all argue against maximizing everything to 254, I do understand why people will want to reach a compromise and come down below that, and in my opinion that is quite reasonable. It is not something that everyone will choose to do, however, and how far down they come is a compromise they should arrive at.


The reason I ask what type of display you have is that I must assume you have a digital display with a fixed and somewhat limited on/off CR range. If you have a CRT, this question isn't really an issue.


Also keep in mind that there may be some trolls who jump in who will disagree with all your assumptions and who will tell you that the world is flat.

 

[krasmuzik]

If enough calibration memories I will do both reference/peak white setups since I do digital front projection that needs every last bit of contrast it can get. Then the customer can go back and forth - or adjust from there for each DVD if they want. If there is headroom I might let it pop cyan above reference white to peak white - depends what they want. For TV or video games I find the bright white scenes and get it maxed without clipping - who knows where it is set at digitally between the broadcaster, the head-end and the settop box.


Since the headroom and footroom serve the same purpose (to have some noise excursion in a digital source) - you would think that one arguing that making BTB visible is the same as making WTW visible. But that is a much more serious contrast hit to raise the black floor for those DVD that don't use BTB for image - so people often will do WTW but not BTB. So it really comes down to the contrast hit as the reason for doing it or not.

 

[ChrisWiggles]



Quote:
Since the headroom and footroom serve the same purpose (to have some noise excursion in a digital source) - you would think that one arguing that making BTB visible is the same as making WTW visible. But that is a much more serious contrast hit to raise the black floor for those DVD that don't use BTB for image - so people often will do WTW but not BTB. So it really comes down to the contrast hit as the reason for doing it or not.

They do not serve the same purpose at all. This is why I refrain from calling values above nominal reference white "whiter than white" which mistakenly implies that it is very much synonomous with "blacker than black." I always try to refer to values above reference white as "peak whites" expressely to distinguish them from values that are below black. Values below black generally are not seen at the display at all. Peak whites, however, absolutely are seen. (for a reference situation). One should never raise black level so as to clearly render values below black on the display, this will cause an artificial elevation of the black level and terribly wash out your image.

 

[Daniel.N]

Thanks guys, I'm very glad to hear that not everybody is shooting for 254, since I tend to shoot for 245 or so, and for exactly the reason you mention - I'm looking for a good tradeoff between maximizing CR and not crushing peak white details.


FWIW, my PJ is a DLP, the Benq 8720 with firmware 1.02 (EU .. 1.01 was equal to the US 1.15 I think). Maybe the PJ is the reason I "dare" drive contrast higher than to shoot for 254, because I really do not notice any color shifting going on. Maybe this is something you got to "teach yourself" to be able to see (??), like many other visual artifacts, or maybe my vision is impaired or maybe the 8720 is .. weird, but at the end of the day, colorshifting really does not affect where I set my contrast at.


If I get some spare time, maybe I'll try to create a program that analyzes screenshots so I can get some hard numbers on how far above 235 there really is data. Or maybe there is an image application out there with better histogram functionality than the ones I'm familiar with.

 

[Daniel.N]



Quote:
Originally Posted by Daniel.N /forum/post/0


maybe there is an image application out there with better histogram functionality than the ones I'm familiar with.

Looked around a little and so far FuturixImager is the best I've found, attaching a screenshot. I just wish it allowed for resizing the histogram and more labels on the horizontal axis.

 

[Daniel.N]

Exifpro has nice histogram functionality in that it allows you to select a part of the image to check, but worthless labels again

.. attaching screenshot.

 

[ChrisWiggles]

Yeah, those labels are strange. The futuriximager has 0-255 labels on both axes, that doesn't even make any sense!