Aller au contenu principal

Setting eyedropper sample size

Commentaires

5 commentaires

  • Ian Wilson
    It’s not currently possible, but it’s a feature that users have been suggesting. I have no knowledge of whether it’s a feature that’s in the pipeline for a future version, but I agree it would be a good one.

    Ian
    0
  • SFA
    [quote="GaryBox" wrote:
    Is it possible to change the sample size of the eyedropper?

    Especially in normalizing skin tones. When picking or applying normalized color, it seems like it is on a single pixel basis, which creates a lot of variation. It would be nice if it sampled a larger area for less variation. I know in Photoshop, you can select a sample size, of point (single pixel), 3x3, 5x5, 11x11, 31x31, 51x51, 101x101. I am thinking even 11x11 would help average out a little better.

    Is this possible?


    When you mention "Normalize" are you thinking of the Normalize tool or the Colour Editor Skin Tone adjustment?

    Normalize is looking for a single colour reading so in some ways need to be rather specific and I would have thought that picking the sample point is mainly a reference that get one to more or less the place from which to tweak the colour values to get provide the calculations that allow the other images to be amended. The detail of colour control is within the tool not the picker.

    For the Skin Tone tool the picker is, again, an approximation and, as with all of the colour tools, selects a range of values form the general area "picked" plus some logical extensions that are usually going to make the adjustments more pleasing.

    The entire range can be adjusted to suit after picking and indeed the picking reference point can be changed without re-picking. Now that is probably easier when expanding rather than contracting BUT more often than not contracting will give just one or two colours (not necessarily contiguous especially when dealing with noise) out of millions and may well lead to adjustments that are so subtle as to be just about invisible at anything less than 400% zoom.

    It may be that for some colour matching work that is a vital requirement but I have to say that given all of the other variables that would be in play I'm struggling to see the reality of such a requirement.

    But the real point is that one can undertake the precision colour (etc.) picking within the tools when necessary and that the picker size, as enacted by C1, may not be as relevant to success as one might imagine. Or at least that is my take on it at the moment but I am happy to learn more about the subject.


    Grant
    0
  • Gary Box
    Grant, that is entirely possible. But not quickly. I use normalize as a FAST way to make skin color match in a lot of images in different locations. It can be a few hundred images a day. Going through skin tone adjustment is ideal on single images but slow and inefficient on large numbers of images in differing lighting and locations. Unless I am using it wrong.

    Normalize does a pretty decent job at this. But if it would average 11x11 pixels on sampling and applying, it would be better. There would be less variation because you clicked on a pore or tiny specular highlight.
    0
  • SFA
    [quote="GaryBox" wrote:
    Grant, that is entirely possible. But not quickly. I use normalize as a FAST way to make skin color match in a lot of images in different locations. It can be a few hundred images a day. Going through skin tone adjustment is ideal on single images but slow and inefficient on large numbers of images in differing lighting and locations. Unless I am using it wrong.

    Normalize does a pretty decent job at this. But if it would average 11x11 pixels on sampling and applying, it would be better. There would be less variation because you clicked on a pore or tiny specular highlight.



    Ok as we are indeed talking normalize I take your point willingly but I'm not sure that the intended effect of normalise is entirely aligned with how you describe your use of it for your purpose.

    I seem to recall a couple of threads some months ago and some other explanations in either tutorials or webinars that discussed the usage and intent so I'll see if I can dig them out.

    That said I would imagine that for your purpose, having re-read the Help information to refresh my memory and check for any additional input I had previously missed, the degree of precision required for Normalize to be most effective might be greater than would be necessary in your situation. I can't be sure of course but the description does stress the benefits of working with very similar lighting.

    The other thing that occurs to me, and perhaps somewhat supports your suggestion, is that the effects of sample size measured in pixels related to subject matter area could be significantly affected by sensor resolution.

    Which thinking leads me to ask whether you find any accuracy benefit from working at, say, 100% view - or even greater zoom levels?

    Come to think of it this subject and a comparison between the Normalize tool use and Skin Tone adjustments in the Color Editor and how to get the best out of them would make a good Tutorial or Webinar subject.

    The skin tone editor is covered from time to time in some types of Webinar but mainly on a single image basis rather than a batch application or with mass processing timing saving considered.

    I recall thinking when I first investigated the Normalize features that moving the picker around in the proposed picking area could produce significantly different colour indications and that precision for matching the selection are and the target areas to apply to were key to full success. Especially across a batch of images. Also that small batches of matching seem more usable (probably obvious) and, as the Help advice suggests, results from work performed early in the process were much more likely to be consistent across a greater number of images at a time.

    I also felt that starting with a Linear curve seemed to be a better way to go. However, as I almost always start with a Linear curve when working with a RAW file that may just have been confirmation bias.

    And then, of course, there is always the potential influence of noise ...

    I would suggest creating a Support Case and making the suggestion for an enhancement. I suspect that there may be some wider (if you will excuse the pun) considerations to go with the ability to change the sample size and thus the calculations for averages.

    Indeed experimenting as I have been typing this leads me so wonder about the potential de-benefits of both precision and reducing precision, however slightly that precision may be reduced (and assuming that the existing selection criteria are fixed AND high precision with no other selection massaging being applied.)

    A most interesting subject.


    Grant
    0
  • SFA
    Looking for something related to a completely different topic I found myself in the newly revamped (and still being developed) FAQ section of the Resource Hub and spotted this article.

    https://support.captureone.com/hc/en-us/articles/360003078197-How-accurate-is-the-White-Balance-picker-


    It occurred to me that in some ways this offers a "sort of" way to emulate the multi-pixel area selection although the amount of adjustment would, presumably, depend on the pixel dimensions of the original file, the pixel size of the preview file (i.e. how small it is) and the current zoom level.

    Of course there are some other issues about what has happened during the 'averaging' as well along with other considerations.

    However, I thought it might be of interest to anyone who has not as yet seen the information.


    Grant
    0

Vous devez vous connecter pour laisser un commentaire.