Photo Match for Professional Architectural Photomontages
-
@unknownuser said:
As plot says the photo-match tool is not very efficient. Don't get me wrong SU itself is fantastic for photomantage.
@unknownuser said:
I much rather have 3DS max's method of placing matching points on the model, then their equal position on the photo and have the application work out the camera position from that.
The method above can actually be duplicated with SU and Corel and probably PS as well.
When I have very difficult mapping to do I take a shot of the piece within SU. I bring it into Corel set it down as a layer. I then creat my material layer over top and cut the material to match the SU image below. I import the texture and apply it....it fits like a glove and if not you can adjust it further by repositioning using the texture positioning tool.
I don't use Photomatch to map textures, but to match the camera so I can render the model in V-Ray and merge the images in Photoshop.
-
As dcauldwell said I think photomatch is really good for smaller buildings. I don't think that 3d Studio or AutoCad are significantly more accurate though when it comes to complex geometry. I tend to just go out with the camera and photograph every elevation I need and then apply them all as textures to an estimated or measured model. it takes longer but it is much more accurate
-
Now I am confused?
Are we talking about using photomatch just to extract textures, or photomatch to display a model in actual context? I thought the OP was talking about the latter. -
Yes, sorry, drifted away from the thread a bit there
-
@unknownuser said:
I would really love to use Sketchup for this photomontaging but am a bit concerned as to its accuracy as it doesnt require camera location, focal lenght etc.
Personally I wouldn't hesitate to use it for small or large buildings.
Although it doesnt require camera location, focal length etc, the technique it uses is equally as good - matching persective.
It gets easier to set up if the photo you are matching is taken at approx 45deg on plan, but clearly this is often not possible - so you make do.
As photographs are often flawed in some way (barrel or pincushion distortion) it still comes down to making final adjustments by eye.
For the record, the process I use is:-
a) on site take several photos, guessing at where you will prefer your viewpoint for the finished montage
b) create the model and match it to the photo in SU
c) render the model (I use vray)
d) export the render (I use png files)
e) Using photoshop (or equiv) paste into a new layer over the photo - sometimes you need to create further layers and copy foreground info so it displays in front of the rendered model.Its dead easy once you've done a couple of times!!
David
-
dcauldwell, amazing image. it took me quite a while to understand which part is not real
-
I found and saved these a while ago, very good show on how to set up perspectives (as tought in the old school days).
Cheers Jeff
-
Cool tuts!
Will be a classic! -
A quick sample of my photo matching, this was done in SU6 in about 1 hr straight from an origonal photo, can give clients a very good ex. of what there addition looks like.
I had to model the letterbox, fence and tree in foreground then locate in photomatched model. The picture probably needs fixing up in photoshop / gimp or similar to take some edges of.
cheers Jeff
-
Thanks for such a useful tut. It really explains perspective rules .
I'm translating it in FrenchMALAISE
-
@dcauldwell said:
Here is an example
(the modelled and vray rendered part is the rear single storey extension)David
So david,
where is your models?
it's seems like photograph for me -
G
Here is the rendered model. It is then merged in with the photograph in PS.
The mirror ball in the foreground was to check out the HDR sky, to see if it was positioned correctly.Cheers David
-
the other problem with photomatch for me, is that is creates a 2 point perspective or some odd setup.. which isnt supported by Maxwell... so its useless... my workaround is to use photomatch..tape a piece of trace to the screen.. trace over a few key edges... and then use that to create my own scene... haha..
-
. . . and if you make a mistake, do you use correction fluid on the screen? I'd love to see you working on a sketch!
I think SU actually creates a 3 point perspective, but because most photos are taken at eye level they look like 2 point.
David
-
haha... yeah, its definitley a half a$$ method..i admit!
i dont know.. they do appear to be 3 point, but i dont know, they are weird.. and Maxwell & Podium dont support them when the output
-
@dcauldwell said:
G
Here is the rendered model. It is then merged in with the photograph in PS.
The mirror ball in the foreground was to check out the HDR sky, to see if it was positioned correctly.
Cheers Davidgeeez, i see, really impressive work
allow me to add this link to video tutorial section.
Thank's
-
Hi idraft,
@idraft said:I found and saved these a while ago, very good show on how to set up perspectives (as tought in the old school days).
Do you happen to know the original author? Your use of the word "found" suggests that it is not you.
Also, ANYONE,
Months ago I ran across a similar tutorial about the geometry of photo matching that I have not been able to find again. (It was saved on my employer's laptop and was in a folder I missed in my last-minute archiving on the day of the layoff. Word to the wise...)
The tutorial I'm looking for, if I recall correctly, used a ranch-style house and constructed the camera position(s) someplace in the front yard. Again IIRC, it took a different approach than the files idraft found. In the above files A-E, the image plane and view line are straight and an object is constructed at an angle and then finally rotated to square. In the tutorial I'm looking for, the object was presumed square with the axes and the technique constructed the image plane at an angle. Also, IIRC, it did not use any actual photos but instead illustrated the geometry with a line drawing instead of a photo.
I suspect that this "missing" tutorial was old, possibly from SU v4, prior to the built-in Photo Match tool.
Does this ring any bells in anyone's memory?
Thanks,
August -
This is where I found them back in 2004 (remember then?)
http://groups.google.com/group/Sketchup-Pro/browse_thread/thread/14168c7bbabd829/bd0617c15f78519dSeems like there are two authors being Michael Young and Ron Kempke (3pt construction). The download of the .skp files is in the link above.
-
Thanks idraft,
@idraft said:This is where I found them back in 2004 ...
The download is in the first message in that thread, which seems to be wrapping up a previous discussion from elsewhere.
The thread includes a very good discussion of field-of-view and how it relates to the angle in the human eye where peripheral vision takes over from the sharper vision of the retina's central area.
It also has an offer from rkempke in October 2005 to put together step-by-step tutorials using camera_parameters.rb, but that was apparently delayed by "solutions for every conceivable situation you may encounter in a 1-pt., 2-pt., or 3-pt. perspective, but each situation requires its own tutorial with a unique set of instructions and calculations."
If anyone knows if Ron ever completed that project, I for one would be VERY interested. That offer was the most recent post from him containing the word "perspective".
Thanks,
August -
@idraft said:
http://groups.google.com/group/Sketchup-Pro/browse_thread/thread/14168c7bbabd829/bd0617c15f78519d
That thread also included an offer from Todd Burch, "I'm sure I can write this in Ruby... it's just a time thing right now."
I suspect all such efforts were put aside when v5 (I think) introduced the PhotoMatch tool.
However, PhotoMatch has its limitations. I'm pursuing this because I'm interested in getting 3D positions of points when you have several photos from different angles that contain the same points, but you have few or no perspective lines/edges at all.
Thanks again,
August
Advertisement