Gradient Tool

Moderator: jsachs

Robert Schleif
Posts: 340
Joined: May 1st, 2009, 8:28 pm

Re: Gradient Tool

Post by Robert Schleif »

It sounds like part of what is going on is the phenomenon known as Mach bands, see Wikipedia, https://en.wikipedia.org/wiki/Mach_bands. It is possible that another part of what is going on is that at the dark end of the intensity range, 8-bits provide barely enough gradations to make the bands indistinguishable, resulting in visible posterization, see http://gene.bio.jhu.edu/gamma_and_perception.pdf.
Marpel
Posts: 692
Joined: September 13th, 2009, 3:19 pm
What is the make/model of your primary camera?: Nikon D810
Location: Port Coquitlam, British Columbia

Re: Gradient Tool

Post by Marpel »

Dieter,

At your suggestion, I did look at the histogram of the B and W version of the gradient and it does indeed look like a rectangle (except for a single spike). And you also made mention of 16 bits towards the end of your second to last post and I think it was in relation to generating a coloured gradient. If that is the case, I did generate a coloured gradient but it is 24 bit. I could find no option to make a 48 (16) bit coloured gradient direct from the tool (without using an 8 bit mask).

Jonathan,

I displayed a saved gradient (black and white generated from PWP) on two monitors of a friend - one an sRGB display and the other an Adobe RGB display. Both showed the same banding (although his Adobe RGB LG did show quite a bit less banding). I have yet to try and change my profile on my NEC (not at home for the weekend).

Robert,

I read through (which doesn't mean I actually understood all I read) the two papers and I do see some relationship between the comments in the articles to some, if not all, of what I see. Thanks for the links.

So, I guess, in spite of what I see on my screen, My biggest concern/question is, if the banding that I see is related to an 8 bit image, I presume they will be visible in print to some extent?
Jonathan - any reason why a coloured gradient straight from the Grad Tool is not 48 bit? Or, in your opinion, is there no/little real difference in this regard?
And if what I see is only related to the monitor, I can presume it will not be visible in print??

Thanks to everyone who responded.

Marv
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gradient Tool

Post by jsachs »

The likely cause of this is small kinks in the monitor calibration curve stored in the monitor profile. Ideally the calibration software would smooth out the curve but many do not. You can either ignore the bands or disable the program that runs on startup to apply the calibration curve and run uncalibrated. In any case the bands should not print as they are not present in the image data.
Jonathan Sachs
Digital Light & Color
Dieter Mayr
Posts: 453
Joined: April 24th, 2009, 11:47 am
What is the make/model of your primary camera?: Nikon D700
Location: Salzburg / Austria

Re: Gradient Tool

Post by Dieter Mayr »

Marv,

If your histogram is ok it means that your data is ok, so the bands are caused by the (monitor) output chain somehow.
If it is possible a printout of a small strip of the image would make clear how the print looks like.
Dieter Mayr
Marpel
Posts: 692
Joined: September 13th, 2009, 3:19 pm
What is the make/model of your primary camera?: Nikon D810
Location: Port Coquitlam, British Columbia

Re: Gradient Tool

Post by Marpel »

OK, thank you everyone. As long as this issue does not show in a print, I can be happy with that (I would consider purchasing a new, pricey monitor but all my spare cash goes directly to buying toys for my grand-daughter!!)

Regards,

Marv
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gradient Tool

Post by jsachs »

Even a pricey new monitor can show this kind of banding if the calibration software does not properly smooth the LUT values it stores in the monitor profile - especially since most sensors are little noisy when reading very dim monitor output. Back when we used to sell a monitor calibrator, I saw this happen on my monitors and added smoothing to our calibration software. Unfortunately, our hardware supplier stopped selling their calibration sensors so we had to abandon the product. It is theoretically possible to extract the calibration curves from the profile, smooth them, and write an updated profile back out - perhaps I'll look into it at some point.
Jonathan Sachs
Digital Light & Color
Robert Schleif
Posts: 340
Joined: May 1st, 2009, 8:28 pm

Re: Gradient Tool

Post by Robert Schleif »

I expect that screen readout tools query the memory that is directly written to screen. Is this in fact so? If so, would the readout tool be helpful in diagnosing the phenomenon? For your purposes, a readout tool that determines intensity at each point along a line would be easier to use than one that provides only point by point information. The free program ImageJ possesses such a tool.
jsachs
Posts: 4203
Joined: January 22nd, 2009, 11:03 pm

Re: Gradient Tool

Post by jsachs »

The readout tool displays the underlying, device-independent image data in the working color space and not values converted to the monitor color space. In addition, the hardware lookup tables built into the display adapter and loaded by the calibration software further modify the image data before it is displayed. Theoretically, the lookup tables are correcting for deviations of the monitor from the ideal response defined by the monitor profile, but if the calibration has a little noise in it the lookup tables actually introduce variations unrelated to the monitor's true response curve and thus some banding shows up. I tested this on my own monitor which I calibrated with an X-Rite Eye One and it also shows a little banding which pretty much goes away if I disable the program that runs on system startup to load the lookup tables and then reboot.
Jonathan Sachs
Digital Light & Color
Post Reply