comparing crops (portrait)

[michapma]

I'm learning what I can about seeing composition. This is no great shot, I especially wish the focus were shifted back a bit. However, I find the curve of the blanket and the natural lighting make the photo somewhat attractive. Since the composition can, as always, be improved, I wanted to work with what is in the photo. The original image and several crops are linked below in reduced size (roughly 115kB per image).

I've maintained a 2:3 ratio in all the crops except one. For me the dominant compositional elements are the line of hands up to face and the curve of the blanket. Unfortunately, due to the focus the nose draws the most attention. I like that crop 3 works well with the rule of thirds, but its drawback is losing the curve of the blanket. Which crops do you like and why, and how would you crop it differently?

original (http://people.ee.ethz.ch/~chapman/images/photos/amber/IMG_5108.jpg)
crop1 (http://people.ee.ethz.ch/~chapman/images/photos/amber/IMG_5108crop1.jpg)
crop 2 (http://people.ee.ethz.ch/~chapman/images/photos/amber/IMG_5108crop2.jpg)
crop 3 (http://people.ee.ethz.ch/~chapman/images/photos/amber/IMG_5108crop3.jpg)
crop 4 (http://people.ee.ethz.ch/~chapman/images/photos/amber/IMG_5108crop4.jpg)

[PhotosGuy]

#1 & 2: I find the sharp graphics distracting & would rather see the focus more toward the far eye.
#3 The bigger crop makes the OOF eye jump out at me, so #4 would be my choice.
It's a reasonably nice shot. If you print it small, like 4X6", everyone should be happy with it. ;-)

[michapma]

I think I like 4 best too. Too bad it's not in a 2:3 ratio (ie, normal 4"x6" print).

Btw, I'd be happy to provide a larger version if anyone would like it for having a go at playing around with it.

[skyphix]

I must say, nothing bugs me more than when things arent in standard printing ratio (in my own photos, I mean). I try to keep my crops in 2:3 but itn ever works. I like 4.

[michapma]

:)

I should figure out how well my color printer works with the right paper.