Lily problem and solution p.156-157

[dpnew]

Never mind--see below.

[Susan S.]

I don't have a problem here with the book technique - I have the original unblurred background layer still visible as Richard does in figure 6.1 - the
transparent area of the blur is then hidden by the show through of the unblurred background - I've used a blur of about 60 and it looks fine. It only looks odd if you turn off the visibility of the original background layer. If you really want to avoid the use of the unblurrred background then there is an alternative workaround to the one that you suggest - before using the gaussian blur filter on the rorchid background, lock the transparency of that layer - this then gives a neat edge to the blurring! In this case I prefer the result in Richard's book. (I'm not quite sure what the gaussian blur algorithm does to the edge pixels in this case, but it gives a paler background result which I like less )

[dpnew]

Ignore my previous post--my mistake

Hi,

Thanks for the response.

lol. I just redid the steps, and the result was terrific. I don't know how many times I did the steps before and got a horrible halo effect all around the orchid. Sheesh! I'll have to see if I can recreate my mistake. My credibility is shot now. My apologies to Richard. (and the Lily title of my post? After all, we're talking about orchids here.)

before using the gaussian blur filter on the rorchid background, lock the transparency of that layer - this then gives a neat edge to the blurring!

I didn't know you could lock the transparency. Thanks!

[edit]
"If you combine the Orchid layer, the drop shadow layer, and the blurred background layer, you'll see a disaster."

I guess my mistake is as plain as day right there: I never had the background layer turned on.
[/edit]

[Richard_Lynch]

Susan, the gaussian blur effects pixels to a distance of about 2.55 times the radius you choose, and does so with some type of bell curve based on gaussian equations...I never really worked out all the math. Usually it suffices to say it works more like a diameter than a radius.

Dpnew, The exercises were gone through many times between me, my excellent technical editor and several people on the editorial staff. More disconcerting than the process of creating the steps was the copy editing, which comes after the writing and introduces errors because of grammatical rules that the CE (who knows words and not the tools) changes things to seem right, but they end up wrong. There are a few instances of this in the book, most of which I believe I caught and have in the errata. For the most part, the solution to any problem you see is taking your time and trying it again.

Sorry I didn't soo this until after you had edited...I might have been able to tell you what you were doing to come up with you result.

[dpnew]

Dpnew, The exercises were gone through many times between me, my excellent technical editor and several people on the editorial staff.

I know it must not be easy, but the mistake I found in the CMYK separations section of the book had me gun shy. I know it wasn't easy for me to spot the error, and I consider it pure luck that I found it. So, after that, when I couldn't get the orchid selection to work properly--even after repeated attempts--I suspected another error.

Even after getting the orchid example to work with Susan S.'s help, there was still something that didn't sit well with me about having to use the original image background layer to come up with the final result. After all, you have the Orchid Background layer and the Orchid layer, and it seems like you should be able to piece them together for the final image. I was just playing around with the example again, and I discovered what was bothering me.

I think the image chosen guarantees a good result. That's because in the original orchid image the background is already blurred. When you apply the gaussian blur to the Orchid Background layer, it exposes a band of transparency all along the Orchid Background border, and it extends inside the Orchid Background border. Adding the Orchid layer and the drop shadow layer does not cover that band of transparency, so the background layer is necessary to fill in that band of transparency along the border. However, if the background wasn't blurred to begin with, it wouldn't look correct.

I know because I tried it by using the bottle.psd image from the saturation masking example for the Orchid Background instead. The bottle image is sharp everywhere. I used the same selection for the Orchid Background layer with the bottle image to create the "Bottle Background" layer with the orchid shape cut out of it. I blurred that layer just like the Orchid Background layer, and then I combined the Orchid layer, the drop shadow layer, and the Bottle Background layer, with the original bottle image background layer. In this case, since the bottle image is sharp to begin with, it shows through fairly sharp in a band all around the orchid, and does not match the gaussian blurred Bottle Background--not a good result.

So, what is really happening with the orchid example is the original background is showing through in a band all the way around the orchid, but because it was blurred to begin with, you can't really notice the difference between that band and the gaussian blurred Orchid Background. Therefore, I don't think the orchid example is very good one for demonstrating how to blur the background. I think a good example shouldn't use the original background image in the final result, and the original image should have a sharp background. I think the example does demonstrate the drop shadow well.

[Richard_Lynch]

I am not exactly sure what you are describing. The idea of the book is advanced work and concept, not to be able to get something to work in a particular image. The key, really, is 'advanced'. While a newer user can get through the exercises, this was not at all intended to be an easy ride...I expect the book to make it easier the second time through (when I have more distance and can read with a new eye).

As far as the 'lilly' is concerned, I am not sure what you mean by there being transparency. The selection around the flowers should not be tremendously soft...if it is, something is wrong. The theory of it is, cut the one part from the other and isolate the terms which you deal with the parts. The blurring should be applied to the background in isolation (try clicking the little Lock Transparent Pixels button on the Layers palette if you are getting too much fringe). If you blur without isolating, the flower will bleed back into the image...and the exercise would completely fail.

The difficulty here in picking an image is to not create a situation where the already long exercises end up extending 4-5 times their length--perhaps becoming tedious as well as redundant, and leading to results that are unrealistic to attain in other images. making a difficult selection can take up many useless paces (at least for the current example we are discussing). if I were to write "make a selection of the XYZ in this very crowded image..." the next question would invariably be 'how?' Sometimes that how can take up many pages, and it would not-at-all be the point of the exercise...and that point might be lost in other steps.

While there may be a few typos in the book (and I cannot but continue to appologize for them, point to the errata [which I believe covers the CMYK issue] and can't exactly vouch for them as well-meaning copy-editing seems to be the bane of good technique--adding as many errors through grammatical concern as it removes, but making them the harder to find), the purpose of the book is CONCEPT. If you are just ending up following through the steps and not attempting to grasp the point, then trying to apply the technique to other images will be impossible.

About the bottles. If you successfully isolate them from the background, you can attempt to blur to create depth-of-field, but the application would be tricky, and/or wrong in that image. Blurring everything in the image beside the bottles will not take into account the difference in distance that they are from the area that is in focus. The fact is you can't willy-nilly slap a drop shadow between a subject and its background and have it look correct: you have to choose the right method for the right image. If the image was shot with intense depth-of-field as the intent (as the orchid was), enhancing that by the methods described works. This is why it is a good image to pick--from my perspective. If depth-of-field for isolation was the intent of the bottle image, the photographer best go back to the drawing board and learn how to take a picture. it isn't that it cannot be done, but just as you want to capture any image as closely as possible to the result you envision, you do NOT want to take on the responsibility of adjusting hundreds of details in an image that could have been dealy with in a moment during the capture. at the same time a very slight drop-shadow (or glow in this case) may give the bottles more presence. this type of adjustment will nearly always need to be on the verge of imperceptible. It would be like taking a shot of a blond person in a light coat against a snowy background in the distance...the blur from controlling the depth-of-field might not have been enough to make good separation, and that can be enhanced. At the same time if the person is holding skis, you would not select the person independently of the skis and blur or you have created an unrealistic effect.

The exercise shows just one of umpteen methods of dealing with separation. You could do something as destructive as the following after separating the parts:

1. duplicate the 'background'
2. blur
3. set to multiply (or darken).
4. group with the 'background' where the orchid has been removed.

This will keep the 'background' solid and enhance with additional layers. again, I can't possibly cover every variation. Combining techniques needs to be the reader's option -- too many combinations, and again I am off the track and point of the exercise.

It is more important to understand why you want to do it. You should have purpose behind creating separation and creating CMYK. Reasons for these would be to help a subject stand out (perhaps using shadowing or blurring) or because you can't do CMYK in Elements with the tools the program gives you and sometimes you may need CMYK, respectively. When you understand the why and what is to be accomplished, this is when you can look at an image with realistic expectations for results and when you will envision what the image could be, rather than what it is. Often the difference is a subtle one, rather than a flagrant change. The technique you use on any one image will probably NOT transfer exactly to another -- it is inherent that you need to adapt and combine technique.

I would much rather take a look at a specific image and try to drive toward a solution and discussion of that (which is moving forward) than look back at what-ifs and could-haves, unless there is a pertinent change to make to an image that we could all use in discussion.

Pick an image and tell me what you want to do and make it available...I think we'll get more from that.

[dpnew]

Hi,

Thanks for the response.

As far as the 'lilly' is concerned

Hey, I changed the title of my post--it just doesn't display correctly.

The theory of it is, cut the one part from the other and isolate the terms which you deal with the parts. The blurring should be applied to the background in isolation...

Yes, I understand that, and that's what I did.

About the bottles.

You misunderstood what I did. My claim is that the orchid image has a blurred background to begin with, which makes the technique succeed, and your technique would not succeed on an image of an Orchid with a sharp background. You can easily verify that. I did by using the bottle.psd image which makes the problem very apparent, but NOT to select the bottles. This is what I said in my post:
---
"The bottle image is sharp everywhere. I used the same selection for the Orchid Background layer with the bottle image to create the "Bottle Background" layer with the orchid shape cut out of it."
---
What I did was: I pretended the bottle image was the background behind the orchids, and I used the orchid mask produced by the channel calculations to select a background from the bottle image. In other words, I cut the shape of the orchids out of the bottle image and used that for my background in the orchid example. Of course the composite image doesn't make sense, but it demonstrated a problem with your technique. The problem is: if you use an image with a sharp background, your technique for blurring the background in order to focus attention on the subject will produce a halo of semi-sharp background all around the subject.

The technique you use on any one image will probably NOT transfer exactly to another

Ok. But, I would hope the techniques you describe would apply to other images and not just the image in the example. Blurring an already blurred background doesn't seem like a very valuable technique to me, although blurring a sharp background to create depth of field does. Personally, I was able to find a workaround--one way is to duplicate the original background image and apply the same gaussian blur to it that you applied to the Orchid Background layer-- but other readers of your book may wonder why your technique doesn't work well on a subject with a sharp background. So, I think discussions like this can provide some answers for other people.

While there may be a few typos in the book (and I cannot but continue to appologize for them, [and] point to the errata [which I believe covers the CMYK issue]

As I pointed out in the CMYK thread, at one point the book says to use the Lighten blend mode, yet the Hidden Power tool uses the Screen blend mode--they produce different results. How much that would effect the actual CMYK printing, I'm not sure since I don't have a printer, and I couldn't test it out. But, it is a mistake in the book, and it is not addressed in the errata.

Pick an image and tell me what you want to do and make it available...I think we'll get more from that.

Well, I do have an image with a blown out sky that also shows through a tree in the near foreground, and I would like to apply a very subtle light blue haze to the sky. I've tried saturation masking and masking with tone to select the sky, but the sky selection around the tree and the parts of the sky that show through the tree doesn't come out well. I've been working on it for about a week now, and I haven't really been very satisfied with any of my results yet. Here is the link:

http://public.fotki.com/dpnew/processing/103_0374.html

[Richard_Lynch]

I don't think you are right in saying the technique doesn't work with other images. No matter what image you use, the technique should work if you do it correctly. You want to blur the background without losing sharpness of a particular object. No matter what the selection, shape of the object, etc. I quickly suggested two minor differences in the last post.

To take your scenario:
1. Complete the selection of the orchid
2. Transfer that selection to the bottle image
3. Convert the bottle image to a layer (no background).
4. cut the selected area.
5. Reselect
6. Paste.

You now have two layers, one with the shape of the background and the other with the shape of the orchid.

7. activate the 'background'
8. click the Lock Transparent Pixels button on the Layers palette.
9. Blur till your heart's content.

There are variations on this where the background can be duplicated, with or without the orchid, and then layer modes utilized to lighten, darken, multiply, define difference...I can't tell you what effect you are trying to achieve.

IN THE EXAMPLE, however, I knew I was going to apply a drop shadow. As the amount of blur was going to be low, the drop shadow more than covers the difference. If you don't like the result, you can affect that by using the solid pixels to your advantage.

Again, the idea: blurring can separate an object. Drop shadow can separate an object. It doesn't matter whether you use a selection of the statue of liberty on a frog...the shape of the result is dependent on your selection. The result itself depends on your handling of the other variables. There are too many variables to cover and/or repeat from what was discussed in the rest of the book. Anyone may have to adjust the technique to achive results in other circumstances.

Getting back to the CMYK thing...whether you use Lighten or Screen, the result will be different, yes. HOWEVER, judging which is BETTER is a different thing. There is not an absolute means of converting RGB to CMYK, and depending on how you want the black to generate, you may do it in EITHER WAY. At the time I wrote the steps, I was interested in using lighten...when I completed the action, I was toying with screen...neither is right or wrong, they are different. If you consider it a mistake that one thing doesn't match the other, I would suggest it is rather a preference. The idea is not to select a screen mode, but to decide how to generate the black. Either screen or lighten can be appropriate. It does not need to be addressed in the errata.

This discussion seems somewhat bent on proving me wrong rather than seeing the point. The fact is there are a lot of ways to accomplish an end. If the technique is whorded and the concept left vacant, the technique is the only thing that gets applied, and the result is either praised or blamed. If you know the concept, the technique can be adjusted to achieve different ends.

Per the image:

1. use Blend Mask to isolate the sky area.
2. create a fill layer with the blue you want to sub in (used a sample from the jeans for blue and did a blue to white gradient).
3. group the fill layer above the blend mask.
4. adjust the opacity/mode of the blend mask to temper the addition.

I am keeping the steps simple here as, again, there will be variations on how you want to approach this depending on how you work. The idea is to isolate the sky and add color to it. As it will be pretty easy to do using a tonal range, Blend Mask will both use the tone source and create the means for applying the color. I only took about a minute to do this...and some of what will be seen as bad selection is actually part of the image (the edge of the brush is blued, which I can't be sure if it is from the exposure or other corrections). To fix this, try cloning in some of the bush color from other sources before adding in the blue.