Log in

No account? Create an account
Another Photoshop entry - Kurt's Life (or lack thereof) [entries|archive|friends|userinfo]
Kurt Onstad

[ userinfo | livejournal userinfo ]
[ archive | journal archive ]

Another Photoshop entry [Jun. 6th, 2004|10:51 am]
Kurt Onstad
[Current Mood |open for criticism]
[Current Music |Seeing Is Believing - Original London Cast - Aspects Of Love]

Here's another Photoshop job from my request for challenges. This one is from self, and you should go read exactly what his challenge was first. Got it? Okay. Now, to see seonsaint and pottertilly on their future honeymoon (pun most definitely intended...),

For this one, the Magic Wand tool wasn't too helpful in isolating the couple, especially where Maria's hair was on a dark background. I made that worse by adding the lighting effect before completely taking out Disneyland, which made everything behind Maria's head even darker, and closer in color to her hair. (Yes, I know I could have gone back, but I wanted to drill into my head that this was the wrong order to do it in, so I punished myself...) So, I went in with the eraser, and manually trimmed out the background. From there, I copied the "couple" layer, and blurred it up. I put the blurred version behind the clear version, and then pulled the eraser tool back out, and erased most of the sharp edges, leaving the blur around the couple, for a (hopefully) more natural transition between them and their new background. The lighting effect I mentioned earlier was a spotlight off to the top right, to make it appear that the two were still actually inside the ship (which is why I didn't mess with proportions at all), but with some outside light spilling onto Matt Saint from the side/rear. But, even with that, things still didn't look quite right, so I adjusted the levels on the background layer, which (to my mind at least), made the two images match more closely.

Okay, people. Now, it's your turn to tell me what I did right, what I did wrong, and all of that other stuff. I know I've got work to do on this, so let me have it. Photoshoppers and laymen alike, I want to hear what you think.

Next on my list of challenges will be shironiku, and then ideaspace (both of which I got in email...). Even though seonsaint sent me his challenge before those two, I'm trying to do these in order from easiest to hardest (I probably should have done shironiku's first, but I haven't found an appropriate background image yet...), so that hopefully, by the time I reach Matt Saint's challenge, I'll have improved to the point where I can actually do a somewhat decent job on it.

[User Picture]From: pottertilly
2004-06-06 11:16 am (UTC)


SeonSaint and I both say "good job"!!! Wow! He said he'll give a more detailed comment later. I am definitely impressed! :)
(Reply) (Thread)
[User Picture]From: seonsaint
2004-06-06 11:54 am (UTC)


Well, first and foremost, as Maria said, good job!

The lighting works great, I think you nailed that pretty well... It very much feels like you have us there in that way.

Biggest problem I think is the selection method you used. I see that you were rightly afraid of 'hard edges' and the very typical cookie cutter look. You did a great job on thinking about how to solve the issue, and I think came very close, I use a very similar technique quite frequently. Basically I think a less soft appearance would be better, I often erase with with the mask layer, creating a mask where I can easily control the edges if I need to. So I suggest maybe experimenting with that.

An older technique I would use involved increasing an images size, doing the cutting in a manner similar to what you did here, and then shrinking it again.

But honestly, very good job, and very good choices.
(Reply) (Thread)
[User Picture]From: self
2004-06-06 02:44 pm (UTC)
There was a trick buried in the two different lighting situations...

The scene's a radiosity nightmare. Nearly everything in that first image is red, and soaked in light. And when you shine a light on something red, you see red bleeding onto everything around that thing. This might not be noticable on their flesh or hair, but their shirts would surely show it.

What you've done with the light values is very good. Meaning, your picture works in black and white. But the colors still need a little tweaking.

There are many ways to do this, most of which undoubtedly better than what I'm about to show you...

My improvised solution, using your image as a starting point:
    Step 1:
    Hue/Saturation.. Drop the saturation to -30. Now, things aren't so vibrant, and the flaws that brought about are diminished. It's still not right, though.

    Step 2:
    Clone the layer your flattened image is on, and go into Hue/Saturation for this new (topmost) layer. Check the "colorize" button. Let's say 40 for hue, 15 for saturation? So, now you've got an aging sepia-toned photo. And it looks pretty good, because again, your light values were right on...

    Step 3:
    Blend this with your toned-down photo by reducing the layer opacity to, say, 31%.

    Step 4:
    Now that their colors are all closer together, let's try to set them back to how things looked in the original picture and see what happens. I'm not comfortable finalizing our layers so far, so I create an adjustment layer of type "levels"... I hit the "auto" button to see what it'll do, and adjust from there (mostly changing the red channel - honestly, I think I only dragged two points).

    My output levels are all 0 and 255 still. My input levels are as follows:

    RED: 7 / 1.14 / 213
    GREEN: 15 / 1.00 / 212
    BLUE: 14 / 1.00 / 209

    ...and, I might take the layer opacity (on this adjustment layer) down to about 89%. Maybe. Can't really make up my mind on that one.

    Step 5:
    The humans in the foreground look a little off now, so let's throw on another hue/saturation adjustment with them in mind. +3/+18, respectively. This actually does a lot for our background action, as well.
Anyway, here's what that ended up looking like:

I'm still not 100% sold on it, but I think the composite's pretty convincing now.

The main thing that I'd like you to pick up from all of that is this:

All of your edits thus far have been made on seperate layers. Which is useful and good, but when you need tight integration, the fastest way to achieve that is with at least one pass on the "final" image as well. (Note that you could have done all of this with adjustment layers and never flattened your image...)

If I were to continue editing from here, I'd make one last change:

    Paste in your original image on top as a new layer, setting it's mode to "color" and it's opacity to 20%.

    This improves the faces but ruins the shirts, so create a mask to make this layer only affect the faces...
(Reply) (Thread)