Replaced clothes and background

[venivedi]

I've edited a snapshot into a photo to be attached on an application paper for admission to a preschool. In fact, I already printed it on 12/13/2003. ^^;

#source :
clothes : a screen grab from a online-shoppingmall
bg : I made a pattern from the bg of the previous clothes, then filled.

Here's the original just resized.
http://myhome.hanafos.com/~venivedi/...m_original.jpg
and WIP
http://myhome.hanafos.com/~venivedi/..._3x4cm_wip.jpg

Any critics and suggestions would be welcome.

[Leah]

My main issue is with the focus - the hair (bangs/fringe, depending on your variety of English) is in focus, then the focus gets softer on the lower half of his face, then the clothing by the neckline, which logically ought to be softer focus still (look at your original) gets sharp focus again, in fact possibly even sharper focus than the hair.

The lower half of his face kind of looks over-retouched, even though I can't figure out exactly what you've done. Maybe it's the ear - did you just copy the ear from the other side? My eye keeps getting drawn to that area.

The background works OK - when you look at it, it's obviously a generated background, because the repeated pattern is quite obvious, but I suspect that wouldn't be so obvious when printed (particularly if you weren't staring at the photo trying to work out what was bugging you about the ear ).

[jch71566]

It appears that you have copied the left half of the face to the right (from the viewing perspective), creating an artificial symmetry. Much of the "character" has been lost -- including the ear!

-Jeff

[venivedi]

Thanks a lot for your input. Actually I've covered highlighted area carefully with right half of the face...to minimize the change of his face contour.
I'll try to fix problems after worktime. Thanks all.

[venivedi]

Am I late a little? ^^
After work, it took about 3 hours for me and my sister to hang papers on the walls of my elder sister's restaurant...with cartoons.

Here's my another try.
# Copy composited image, then apply Gaussian Blur 1px
# Add oval shaped layer mask around face, adjust Advanced Layerblending option.
# Liquify the lower part of the left ear.
# Make RGB channel of the original into a alpha channel(Levels adjusted and apply Gaussian Blur)...copy it into the retouched photo...and its Levels adjusted with that layer-mask.

p.s. In previous attached images,
# I just copied nostrill part to remove snivel...and ear part to remove washed out area.
Other highlight part of the left half of his face just covered with other half of his face...of course...using layermask to restore skin texture. I tried not to change his face contours.
# Liquify : lips and teeth...that's about it. # Blur : Clothes and hair a little

[Leah]

It still just doesn't look realistic to me. I think you would get a better result staying closer to the original. I've been playing around with your source images and have an example of what I mean if you don't mind my posting it?

[venivedi]

It still just doesn't look realistic to me. I think you would get a better result staying closer to the original. I've been playing around with your source images and have an example of what I mean if you don't mind my posting it?

I don't mind. Feel free to post.

[Leah]

OK... I'm not suggesting that this is in any way perfect, and I haven't spent nearly enough time on it, but it's closer to the way I would go about it.

[venivedi]

Leah, I really appreciate your effort. I'll need your suggestion. Thank you.

[jch71566]

OK... I'm not suggesting that this is in any way perfect, and I haven't spent nearly enough time on it, but it's closer to the way I would go about it.

Leah, please share how you did this.

-Jeff

[Leah]

I had a horrible feeling someone was going to ask me to remember this...

The general principles were

• to leave the face untouched as much as possible, since it was such a charming shot (if you ignored the snot...)
• to leave the sense of directional lighting intact even though the extreme highlights needed to be toned down
• to soften the focus on the clothing so that it wasn't sharper than the face and hair
• to try to bring a sense of uniformity in in the colours
• while I was at it, to make the background less obviously repeated.

and the general steps involved, so far as I remember them...

• took the existing background repeat, made it a larger repeat and got rid of the elements that (to me) made it look obviously repeating (I may do a quick tutorial on creating patterns as the Photoshop Wow! book I have explains it so well). Refilled background.
• cut out face from informal shot using quick mask (using the smudge tool on the quick mask where there were tufty bits of hair). Positioned on new image, adjusted size, masked where necessary (plus I think I had to clone a bit of the neckline of the sweater to fill in because I had the head and neck a bit smaller overall than in venivedi's version)
• copied some bits of skin from the shaded side of his face over the blown-out highlights (I think for the bit on his ear I had to adjust the levels to make the copied bit lighter, too). Then Hide All masked them out and painted back in very gently where strictly necessary - low opacity brush, and a small brush on the ear - working very carefully and remembering the general principle that I just wanted to tone down the highlights, not remove them entirely. Also copied some of the hair from the other side over the strong hair highlight, Hide All masked it and then brushed in just a little at low opacity with a soft brush.
• cloned out the snot and the dribble on chin
• I think I did a tiny tiny Levels increase of contrast on the irises and pupils to make the highlights slightly more obvious (though they're still not very obvious)
• gaussian blurred the sweater, in two goes so far as I remember for more and less blurred areas. Did some dodging and burning to darken some areas of sweater and lighten others to tie in with light direction on face. Also darkened very slightly round edge of shadowed ear and that side of hair, I think.
• played around with the color, although I can't remember quite how. I know that one thing I did was to sample a color (part of the skintone, I think) and add a fill layer in Color mode at very very low opacity over the whole image to try to tie everything together, but I think I'd dont somw tweaking beforehand - basically just looking for areas that didn't seem to tie in with the rest of the image and trying to adjust their color to be more in keeping with the whole. None of this was big changes at all, it was all teeny tiny subtle stuff and probably wasn't really necessary.
• dodging and burning again, very slightly darkened background on shadow side of image, again to tie in with the direction of light on the face.

Things that are still definitely wrong:

• where the bottom of his neck joins the sweater - distinct oopsie there
• I still don't think the differential focus is quite right on the sweater - if this were my image I'd play around with it more
• still think it lacks a certain cohesion - if I were doing this again I think I'd duplicate the whole when I got to this stage, run it through Xero's Radiance filter at a suitable setting, then layer back over this at around 25-35% opacity so that it all ties together. That would also give it a bit more "punch".