Steps required in making a Photomontage?

[Anonymous]

Without using the Align tool can anyone sugest a different way of creating a accurate photomontage.

Place a camera on the archicad floor plan in the known position that photo was taken.

Get the center point of the photo a make the target point of the camera line up. ( Does the distance the target is away from the camera make a big difference to the end 3d image ? )

This next bit am not too sure on.

Camera z =
If the gf plan of the model is at level 0.0 &
The photo was taken take from the public road road level being 2m lower than the proposed gf level of building. Taking 1.8m as eye level. Is it correct
to say that Camera z = -200

Target z =
The center point of the photo lands on a point in the site approx 400mm over the proposed gf level of building.
Target z = 400

Change the sun direction in archicad to match that of the photo.

Change the view cone angle until all of the building is in view.

Render the image at a high resolution.

Bring image into photoshop

Scale image to suit image by ;
Do a line section in archicad of site. See where proposed ridge line would hit neighbouring building in background etc.

A bit of fudging croping & cloning in photoshop.

And you get a fairly accurate looking photorendering.

Basically am looking for any advice tips, methods or tricks anyone has regarding same to get a more accurate photomontage.

I have tried the align tool a good few times before but get very mixed results from it.

Any help is greatly appreciated.

[Dwight]

You are right on track: inserting a model into a photo is subjective and requires visual judgement.

The Align Tool is imperfect because photos are imperfect - any lens distortion destroys measuring. Furthermore, most context photos lack sufficient reference points to accurately position the camera. Should Archicad users wish to practice with the Align Tool, i suggest making a rendering of a space with tall walls that fill the scene and then place a piece of furniture or something into that scene. Insertions into Archicad's distortion-free renderings are easy.

When there are poor reference points, I use the parallax method - like the old printed perspective grid sheets with converging vanishing lines popular before most of you were born. Use adjacent building edges to create guidelines in the background scene. Make a stand-in block for the building to simplify convergence.

Also, ground planes are never flat - understanding slight level changes are critical for success. Anything i have ever done requires jinking around in OpenGL for fine tuning.

[Anonymous]

Also realise that if you have to use camera "roll" to correctly align in 3D window, results may very in Lightworks rendering, as camera roll is not allowed for in LW (unless it has been fixed in AC11? Dwight?)

[Anonymous]

DerryP, try this for helping to align your view point:

Set your target photo as the rendering background for your 3d window, then view your model in wireframe superimposed on its actual background image.

This works wonders for lining up those reference planes Dwight was referring to.

The key is to create an unscaled version of the background image that fits on your monitor, and in your 3d window, without distorting or stretching. If you have a high res photo, resample it with same proportions to something small enough to fit on screen for the 3d background, then set your 3d window size to that exact pixel dimension while aligning the viewpoint. When you all aligned and ready to render, set the photorendering window size to the original, high res pixel size (if you can).

You can then render with or without the backgound turned on. Having it on makes your 'blue screen' work very quick and easy.

By the way, you actually set the photo rendering background, then tell the 3d window to use 'same as photo rendering'.

I you have AC11, this is all great because you can set up different views that have different backgrounds and cameras settings all lined up. I don't actually place cameras, just modify the 3d view camera settings and update my views if I want to tweak the viewpoint.

One other thing - remember that camera target height is hard to determine in actual photos. If the camera is at zero, a target height of 5 at 10 distance would give the same apparent center of picture as a target height of 50 at 100 distance, just different field of view.

[Anonymous]

s2art wrote:
Also realise that if you have to use camera "roll" to correctly align in 3D window, results may very in Lightworks rendering, as camera roll is not allowed for in LW (unless it has been fixed in AC11? Dwight?)

Is that a fact?....
Camera roll is not allowed in LW for AC 10?
Yeah, what about AC 11......no camera roll too in LW?
Any confirmation on this?
Thanks in advance for those who can verify in AC 10 and AC 11....

[Dwight]

Putting camera roll into the camera view is replicated in Litewurx.

[Dwight]

There are several relevant points in placing a rendered model into a photo that make Photoshop a necessity :

- this is largely a "manual" task - You can spend lots of time trying to make it perfect at the rendering stage when it is just a few seconds work in Photoshop.

- that usually the photo will have foreground elements that must be separated out from the background through layer duplication and masking or erasure.

- that the final adjustment of the rendering is done using transformation in Photoshop to match camera distortion. The "Warp" tool is quite appropriate.

[Anonymous]

If you need additional help with Photoshop there are these tutorials...

http://laughingsquid.com/you-suck-at-photoshop-by-donnie-hoyle/

I guess you have to figure out what [censored] means...

[Brad Elliott]

The beautiful thing is that the link is still good. It just censored but did not actually remove the word.