transfer a shading
well, this thread is somewhat related to the forest2city thread, but i didnt want to jam it even more with yet another issue ....
Say i have two images, A and B. (see attachment).
A is is a piece of a satellite pic, showing some shaded hillsides.
(from straight up, and from far)
B is a pic of another sat pic, a flat area, and has a much higher resolution
(like 10x better, but not closer)
Id like to assume for a second that B shows the the same area.
And id like to steal the **hillshading** from A and get it into B.
I tried to simply blend the two layers, but i wasnt too happy with that.
What other options, besides layer transparency, do i have to 'transfer' the shading-structure of A, (and anything else that can help to suggest elevation etc.) to B ?
If you think thats a lke a crazy idea and should just forget about it, im open for that as well, actually ...
small pics and B is very hard to see anything in.
however, regardless of that, you can 'set' your clone tool on A and then click on B and clone directly from A to B. or, you could do cut and pastes from A to B. or, you could put A on as a layer in B and simply erase out the parts of B that you want to have the parts of A show up.
If you just want the shading you could try a luminosity selection and then invert. You could then cut and paste the selection to the other photo. Of course there are also blend modes (overlay, darken, multiply...)
a more practical example in a single screenshot ....
A has been blown up to suite the resolution of B, and serves as a guide for pasting.
Top right corner is already covered wth B textures.
What is lost, however, is the hillshading of A in the overpasted area, which originally had a shading pattern very similar to the neighboring area in the upper left.
Hopefully, the manipulated pic, beeing about 10x the size of the original will allow me to zoom in much closer in 3D visualization of that landscape, besides of opening possibilities for landscape reconstruction.
I gave this a try use the blending modes that goose mentioned.
What I did was take your last example and duplicated the layer. Desaturated it to make it grayscale and then applied a levels adjustment to brighten it up a little.
Than I brought in image B and positioned it at the top left next to your example for comparison. I changed the layer blending mode to Hard Light and made it part of a clipping mask by ALT+clicking on the line between the layers. By making it part of a clipping group, image B will only appear where pixels below it are not transparent.
I applied a layer mask to the grayscale layer and using black, painted out the rivers. Since image B is part of a clipping group, the rivers were automatically cut away there, as well.
As the result was a little dark, I applied an adjustment layer to brighten it up a bit and get it close to image B's original color. I also made the adjustment layer part of the clipping group, so it only affected the visible pixels of image B.
The results are not too bad, I think. But, here's the thing... I don't think you'll have much luck using this method to increase the overall resolution of the image. The new shading on image B is still based on the resolution of the original image, which will still deteriorate in quality when the image is resampled.
ok, i tried to follow what you guys were suggesting.
youre right when noticing that the hi-resolution material looses a bit too much of its structure in this process, which perhaps makes trading structure for shading seem like not such a good deal ....
the objective would be to find a compromise and find out how to smuggle *some* hillshading in, taking away only a minimum of Bs resolution.
so i tried tweaking around, combining the things I know how to do with your ideas.
as for a luminosity selection, i didnt manage to do that.
(perhaps thats due to the portuguese keyboard here,
where the ~ char works kinda differently.)
isnt Ctrl+Alt+~ the command for luminosity selection ?
to start, i loaded the example, selected a bit of the already overpasted
area in the top left, copied it and placed it over the top right.
Ill call this B2, ok ?
(This gave me better referenece of what happens to the hi-res material
in the process.)
I selected the area where i intended to steal the shading.
I made a copy of that and desaturated it
I made a selection of its dark colours.
I used that selection to copy parts of B2 onto a new layer.
I darkened the new layer a bit and placed it above B2.
I selected the area where i intended to steal the shading.
I made a copy of that.
I tweaked the copy's colorbalance and contrast into an extreme.
I placed it under B2, which i made slightly transparent (92% op.)
The results are attached.
Maybe you guys have some ideas how i can refine these doings ...
Last edited by doonee; 08-24-2005 at 08:06 AM.
I'm not quite sure I follow what you did. Can you post a screen shot of your layers palette?
Also, it comes to mind that this might be a good place to use some of those DEM images. Part of the loss of the high res detail is because the different lightness values of individual trees in the canopy of the shading source image is adding unwanted noise to the higher res target image. Thus, the details are getting lost. That's in addition to the fact that the shading information is coming from a lower resolution image. The grayscale DEM images (assuming the POV is from overhead) would be a much better source for the shading information.
screenshots of the layers are attached.
Its nothing complicated really, its only that im not doing too well yet in explaining what im doing.
As for the DEMs, im afraid that the ones which are available for this particular area do not have a resolution which helps. Generally, they fail to pick up some of the hills, which is how this all started. (Resoltuion of the the USGS DEM in this case is 1 heightsample/90m)
I found that a slight hint of a shading in the overlay, combined with that of the DEM used in the representation, gives some nice results. Although the POV
is orthographic overhead, usually all sat-overlays come with some kind of a shading of their own, and it was that i was trying to 'fake'.
Last edited by doonee; 08-24-2005 at 08:38 AM.
I see what you did now. It's kind of similar to what I did. In the example I posted, the effect could probably be reduced by lowering the opacity of the grayscale layer, or even better... by not doing as much of a levels adjustment on the grayscale layer so it doesn't get as bright. That way much less noise would make it into B.
|Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)|
|Thread||Thread Starter||Forum||Replies||Last Post|
|File Transfer Program?||Vasquez||Software||4||05-03-2007 07:59 AM|
|Photo transfer to a wall||skydog||Image Help||10||02-21-2007 09:04 PM|
|Advanced color shading - how the hell did he do that?:-)||Faust||Photo Retouching||53||11-15-2006 07:40 PM|
|Software license transfer||Doug Nelson||Software||2||09-16-2006 06:32 PM|
|Polaroid Transfer Effect||Andrew B.||Photo-Based Art||3||02-23-2003 02:10 PM|