Untitled Document

For anyone who's interested, (hoping of course I'm not stating the obvious or covering previously discussed stuff) I have come across something which some may find useful and wanted to share it with the group.

Regarding darkening skies, where Dan has suggested in the past that adding a luminosity layer of the Red channel works well, but has the disadvantage of adding unpleasant noise. In experimenting with his "Bridge of Sighs" Shadow/Highlight technique, I have found that using the Blue channel with this method is a good alternative.

Begin with three identical layers, the top set to Darken and the middle set to Overlay. To the middle layer, Apply the Blue channel inverted, then blur. You have to be more careful with the blur than usual, as the darkening is so dramatic that the halo can be obvious. For this reason, a smaller blur is usually in order and the opacity of the layer can be reduced, of course.

The fact that the blue is so light in skies makes the darkening very effective and although the Blue is usually the worst of the three quality-wise, light skies seem to be less problematic and the blurring takes care of any existing noise or rubbish.

As an added bonus, if the shot is a landscape including foliage, reducing the opacity of the top, Darken layer has a similarly dramatic lightening effect there, as the Blue channel is so dark in these areas.

Russ Brown

___________________________________________________________________________ .

Russ,

I tried this with a random photo where I wanted to darken the sky:

http://jjbong.smugmug.com/gallery/3647167/1#241165293

I wasn't particularly happy with the results, compared with Dan's R channel luminosity method (ignoring the lens problem in the upper right, which can be fixed easily regardless of how you try to make the sky more interesting).

I don't think you want to darken the sky so much as get a lot more contrast in the sky (meaning darkening parts of it).

Looking at the channels in the original photo, I see a lot more contrast in the R (L values from 56 to 18, again ignoring the upper right), while the B channel shows much lower variation (56 or so to only 43).

You're using a different method, and perhaps Overlay mode magnifies the difference somewhat. Although this is a sample size of one, I don't think it particularly unusual.

Also, isn't the sky typically noisier in the B channel on a digital shot? Idon't claim expertise here, just remembering a discussion in PP5E on this (I'm on the road and don't have access to it).

Am I missing something?

John Bongiovanni

___________________________________________________________________________ .

Hi John,

Thanks for the comment. I guess I felt that this method would make a good alternative to the Red blend, rather than a substitute, especially where the Red is very noisy, as it often is in skies. I've used this method quite a few times and been happy with the results. In my experience, as I said in my original post, I have found that although the Blue is usually noisy generally, in skies it is often better than the Red and is further improved by the blur applied to the Overlay layer.

Russ B

___________________________________________________________________________ .

You have given me a marvelous eureka moment Russ - with your 3 layer technique, suddenly the power and combination of layers and channel blending has truely popped into place and I have definitely taken a step forward in understanding what is possible with layer combinations.

However I can not really see why the blur is necessary, as even a small blur does not seem to improve on the blending and, as you say, halos begin to bite - could somebody explain the logic behind the blur please.

Thank you

Chris Broadhurst

Web Site: http://www.broadhurst-family.co.uk/lefteye/

___________________________________________________________________________

Excellent! Thanks Chris, glad to hear it has helped.

As for the blur, it is explained in detail in Dan's PP5E, and although I have read through it many times, my understanding can better be explained as more conceptual rather than technical. I'm therefore anxious about trying to explain it myself and hope that someone more knowledgeable (I'm looking at YOU Stephen!) will put their hand up,(in the absence of Dan himself.) However, briefly, without a blur, the overlay can cancel out detail, leaving the image flat and in many cases, just plain ugly. The blur halo gives the impression of depth and works the magic of the Overlay blend mode. I'm going to put my foot in it if I try to do any better than that! If you have PP5E, go straight to chapter 18, do not pass Go, do not collect $200.

Russ B

___________________________________________________________________________

"Russell Brown" wrote:

I don't claim to be even remotely as knowledgeable as either, but here's my attempt at an explanation.

With Overlay blends, you're using a layer to darken and lighten parts of the image. Russel is using the B channel inverted to lighten where the B is dark and darken where the B is light. Without a blur, this would drive things to neutral in the resulting B channel (and possibly others). You can see this effect by taking a photo with a sky you want to darken, and use the first part of Russel's technique (Overlay with inverted B, but not the darken layer). Don't blur, and look at just the B channel. You'll see considerable loss of detail in clouds, as the pixel-by-pixel move drives everything to neutral.

Put a blur in and see the difference, first in the B channel only (then in the composite). What happens is that the darkest parts of the channel are still being lightened, as is anything nearby that isn't necessarily dark (nearby being a function of the blur radius). This preserves contrast (hence detail) locally.

The same thing is happening in Dan's alternative to Shadow/Highlight in PP5E. You use a channel or channel blend to drive the lightening and darkening, but you use a blur so as not to drive that blend to neutral. If you're lightening or darkening some extreme point, you're also doing it nearby, although not as much.

John Bongiovanni

___________________________________________________________________________

.

Just to clarify, in case of confusion, I was referring to the Blue channel of RGB, which I think is what John means here when he writes "B", as opposed to the B channel of LAB.

Russ B

___________________________________________________________________________

John Bongiovanni wrote:

I agree John, well said. One can also optionally lighten the sky, for more contrast (results obviously depend on the image at hand and the intended edit/mood that one is looking for).

Russell Brown wrote:

Thanks Russell, but you will have to wait for somebody else, when it comes to Photoshop, I currently have Luminosity and Chromaticity on my mind (plus a few other things).

I have some general thoughts on the matter of enhancing sky.

Often one will isolate the sky with blend if sliders, layer masks (gradient or channel based) or combinations of both masks and blend if. Or, sometimes a blending mode for the sky and a rough layer mask for lesser areas will do the trick of isolating and blending the edits over the original.

One will often reduce both chroma and luminosity noise in the sky, perhaps aggressively so. This may be done as a second render if working from raw camera data, then blended over the "regular" render through previously mentioned blending methods. The noise reduction may be done with native tools in the raw processor, Photoshop or with third party applications. One would apply the smoothing before adjusting the sky for "drama".

In addition to the method currently being discussed, there are two other faster options, over and above more regular methods such as curves. As recently mentioned, the CS3 Black & White adjustment command when blended/faded to luminosity mode can affect the tone of chosen hues while preserving others with no/little need for masking or isolating the edits. Another method is to use a variation of HiRaLoAm USM or high pass filtering, running the amount higher and with extreme radius over the normal application of this technique. One may wish to use a "bell curve" layer mask or blend if sliders to restrict the effect from the shadows (helpful for breaking sky from horizon/trees/cityscape etc) and so as to not blow out detail in clouds (if important, I am a fan of highlight detail).

One last thing, no matter what method is used, one could also try placing a Luminosity "Channel" Layer between the corrections and the original image, clipping/grouping the masked sky edits to the Luminosity "Channel" layer (perhaps adjusting it's opacity or blend if).

Sincerely,

Stephen Marsh.

___________________________________________________________________________

.

"Stephen Marsh" wrote:

Stephen-

I've been puzzling over this paragraph for several days, and I have to admit being completely clueless. For example, I couldn't make the leap from the recent discussion on the CS3 B&W adjustment command to its application to darkening/increasing contrast in skies.

I am similarly clueless wrt the rest of the paragraph. Could you elaborate?

John Bongiovanni

___________________________________________________________________________

John Bongiovanni wrote:

The key is blending the CS3 B+W command in *luminosity* blend mode (or clipping/grouping the adjustment layer to a true monotone Luminosity "Channel" layer as mentioned in my recent posts on the accepted wisdom of luminosity editing in RGB). The command will not make a black and white image, but in luminosity mode it will allow one to make tonal edits to various hue ranges in an interactive fashion (be aware that banding or posterization may be introduced with some images and edits). As this is a selective command, one may not have to introduce complex masks or blend if to isolate the edits to the sky.

The second method mentioned simply uses a dupe of the sky, often isolated from the rest of the picture in a layer mask. One applies a rather high amount USM at a very large radius - even maximum radius to the image data. This can also be done with high pass or even with the good old "contrast masking" method that was popular before the Shadow/Highlight command was introduced. It is basically just making the contrast greater, as can be done with curves or other methods. This will often make the sky near the horizon lighter and the sky near the top of the image darker, sort of like a polarizer filter.

The last section was noting that one may wish to concentrate these contrast edits more to the midtones of the image than to the endpoints. Most commonly this is done with a "bell curve" or "endpoint" or "midtone" layer mask. Similar results can also be achieved using blend if sliders, in instead of or in combination with the layer mask. I often like to have more contrast in the sky and part of this is with brighter clouds, but not at the expense of highlight detail, so the lessening of the effect in the highlights may be required.

Of course, if one clips/groups these various sky adjustments to a monotone Luminosity "Channel" layer above the original layer, then the result is less saturated with no hue shift, which may or may not provide the mood that one is looking for (if not, try reducing opacity or blend if of the monotone layer).

John, is this a better explanation of my original paragraph, does it help?

Sincerely,

Stephen Marsh.

___________________________________________________________________________

As I usually have the curve axis set-up for ink rather than light, it is more often than not a "U" curve grayscale mask:

http: //members.ozemail.com.au/~binaryfx/images/endpointmask_bfx3.gif

Regards,

Stephen Marsh.