Compositing 3D into a still image - help plz

# 1 11-02-2011 , 10:44 PM

Subscriber
Join Date: Jan 2004

Location: Bolton - UK

Posts: 2,457

Compositing 3D into a still image - help plz

It may have been a good thing I was away for the last challenge as I have no idea how to do this.

As some know I went to Escape studios today in London to enroll on a VFX class starting in May. One thing the tutor told me to work on is getting 3d in to real images. Now I know what I want to do - put some thing in to my street as during the day its empty and I think there are some cool area to take photos of for a back plate.

I think it goes

1) Take photo`s for back plate and ref
2) Model the object
3) create some proxy models for stuff in photo (eg floor)
4) Line up object in photo (not sure how)
5) Match the lighting (not sure how)
6) Use render layers to create realism
7) Comp together

Its the bits in red I have no idea how to do -

For part 4 I know how to do this for video (matchmoving) but how for a still - is it just eyeballing it.
For part 5 - I know how to make panoramic photos (full 360) but not hdr photos. Our they really needed
I do have a large collection of hdri images - they are just not off my street.

Basically I need help on what sort of photos do I need to take for reflections and lighting - and how (if a hdri) - after the last challenge I know some of you know how to do this...

Quote

# 2 11-02-2011 , 11:15 PM

Subscriber
Join Date: Jan 2010

Location: Budapest, Hungary

Posts: 244

Hey Tweety!

If you re using a comp program (like nuke) make a delens on the picture you took,(using your camera's focal lenght). So your picture wont bend in perspective axis (if i am telling right user added image

)

This 2steps re really simple. Put your photo into an image plane on the Perspective view. Than start to move your camera still the grid fit you pictures ground/table/house whatever you see. than you have the right axis

The light stuff is a bit complicated. You should add ligths like they were outside the place. 1 directional light for the sun & some other lights. I took a picture from my desk & i used 1 direction for sunlight & an arealight as a monitor. If it does not match perfecty you can tuune them in compositing program.

It was a good free friday tutorial.... ..... .... found it finally

https://simplymaya.com/forum/showthre...ht=free+friday

It is just what you need user added image

Quote

# 3 11-02-2011 , 11:37 PM

Subscriber
Join Date: Jan 2004

Location: Bolton - UK

Posts: 2,457

You know what thats perfect thank you - Had no ideas this place was doing this free friday tutorial so double win.

That does leave my questions about setting up the image based lighting and setting up the reflections but its a big step in the right direction... Thank you

Last edited by tweetytunes; 11-02-2011 at 11:47 PM.

Quote

# 4 11-02-2011 , 11:40 PM

Subscriber
Join Date: Jan 2010

Location: Budapest, Hungary

Posts: 244

i wanted to post a smiley.... but thats shorter than 10 character. so: user added image

Quote

# 5 12-02-2011 , 02:22 AM

The Nurb Herd
Join Date: Oct 2007

Location: London

Posts: 2,381

1) Take photo`s for back plate and ref
2) Model the object
3) create some proxy models for stuff in photo (eg floor)
4) Line up object in photo (not sure how)
5) Match the lighting (not sure how)
6) Use render layers to create realism
7) Comp together

Originally Posted by tweetytunes

hey man, seems like you liked the sound of the course.

I think you have the order wrong tho. 4 should come after 3 and you should have a 2.1) create and match camera to backplate.
you can do this pretty easily if you are taking the photo yourself, as you will know (or can find out) the focal length. so create a camera and adjust the focal length to match. then dolly the camera until you think it lines up. Look at my wip for the comp, it might give you some ideas as to how to match the camera.

as for lighting, i would shy away from final gather if thats what you are thinking. try to light it manually. Final gather is not bad, its just not used as much as you might think. It causes a lot of artifacts in your lighting.
You do not need a chrome ball and an 8mm lens to make an hdri. What would be a good cheat would be to take a photo with as wide a lens as you have. if you dont have a tripod then just take one. but if you do, take one at the fstop you need to. Then take 2 up an exposure and 2 down. (hope that made sense. been at the pub)
if you dont have a tripod, take a photo and create the duplicate exposures in photoshop. then use the "merge to hdri" button and it will create an hdri.
you can use this as a reflection environment.

can't think straight so will review this tomorrow. too many beers.
night.

Quote

# 6 13-02-2011 , 05:24 PM

Subscriber
Join Date: Jan 2004

Location: Bolton - UK

Posts: 2,457

Sorry Dom completely missed this reply. Thanks 4 the tips but you see this is where I get confused with the photos. I have a good digital camera 14mp but no lens, no tripod

So if I follow what you say I ...

1) I go out side - take the photo of the area I want to place the subject in.
2) No tripod so I take one photo as wide as I can - but of what - the stuff in front of subject spot or the stuff behind. And I got a cool panoramic mode on this camera and it really does work - is that worth doing.
3) Take photo in to photoshop - And how do you duplicate exposures in photoshop and this "merge to hdri" button.

Many thanks again

Quote

# 7 13-02-2011 , 05:54 PM

Head of 3d Admedia
Join Date: May 2007

Location: nairobi , ken-ya

Posts: 1,340

1) I go out side - take the photo of the area I want to place the subject in.
2) No tripod so I take one photo as wide as I can - but of what - the stuff in front of subject spot or the stuff behind. And I got a cool panoramic mode on this camera and it really does work - is that worth doing.
3) Take photo in to photoshop - And how do you duplicate exposures in photoshop and this "merge to hdri" button.

Originally Posted by

hey check out this link

https://www.talkingtree.com/blog/inde...ake-hdr-images

Quote

# 8 15-02-2011 , 09:27 AM

Moderator
Join Date: May 2003

Location: London

Posts: 1,001

Really is much easier if you have a tripod to do stuff like this.
Ill do a quite detailed reply, its quite useful to know this kinda stuff:

In terms of lining up your set nicely, get some measurements of things to build the set with.

If you have a grasp of the dimentions of things or atleast if you have a few things that have a definite size.
e.g. if you have a building, grab its width and height then you can use that to make a cube in maya to the same dimentions you can then be certain that you line up is accurate. Also try and measure things in relationship to eachother, so if you have a bench infront of a house measure the dimentions of both as best you can and then measure the distance between the two in x and z.

The panoramic mode on the camera is fine but faking the dynamic range in photoshop isn't ideal, if you can afford or borrow a cheap tripod i recomend doing that. Then you need to clamp your exposures to get a series in the range you need.

If your camera has the features needed do the following:
-decide on an apature to use, it wont really matter that much for general purpouses but remember the higher the apature the greater the depth of field (sharper thoughout depth). This needs to stay the same for each photograph.
-Then you need to vary the shutter speed by a stop for each position you are taking to aquire a panorama.
-so for example :
Position one take a photograph at f. 22 at 25 50 and 100/th of a second speeds (depending on the lighting, just if your camera can support changing frame speed on exposures thats ideal but you can do it by eye too)
exposures
Then tilt the camera up and do the same, three photographs- down and repeat, then rotate the camera across and do the same until you get full coverage.

Also its best to have these images as RAW, this will give you better quality down the line.

You then need to stitch and merge the photographs, autodesk stitches does this nicely if i remember, so it does nice things like stack the 3 exposures for you and treat them as a single image and then automatically stitch it togeather.

Once you have the exr or hdr exported from your stitching software (exr is industry standard) you will probably have black areas where you had no photographs, usually at the top and bottom.

You need to fill these areas or they will appear in your reflections and you can get bad lighting from the sky, from memory CS4 handles HDR very badly (clamped high values) but this might have changed in updates. Use nuke or another package to clone brush the dark areas so they are filled.

Be wary that you are dealing with a hdr image and its not the same as a jpg, the gamma curve IS NOT baked in, so for this image to work correctly in lighting you cannot compress this image.

This subject is also one of the most overlooked in CGI, its worth looking into:
Gamma correction
AWESOME link:
linier spacehttps://www.mymentalray.com/wiki/inde...ar_color_space
But basically, all images need to be in linier space to work correctly, so for a jpg texture to appear correctly a gamma correct of .45 needs to be applied to it (can be done in hypershade with a gamma correct node) this removes the standard baked in 2.2 gamma correction from the image (so it looks nice for our eyes) so its mathematically correct.

This is all the more important for hdr, as its going to be lighting you scene etc.

If you want to see how an image properly you can use a LUT in your compositing software to apply a gamma to your display without damaging your composite.

Its not over yet!
To have this work nicely you need to do two more things:
Neutralize the image ( this is so the lighting is predictable)
Create a range of blurred/ scaled images for use in lighting / reflections.

To neturalize an image its best to take a picture of something thats 50% gray at the time you where taking your hdr photographs, a macbeth chart is ideal. Here. Because you have a range of colors (or atleast grays) you can be sure of you can then tweak the color balance of your image to be normal, and more importantly use theis to perfectly tweak your IBL settings in maya to a gray sphere until it matches the middle gray part of the macbeth chart perfectly

Output 5 ish images of your hdr that are quite blurry to practically totally diffuse, this is for lighthing, you won't need sharp or large images to get nice results.

For reflections perhaps go for 2k images that range from very sharp to relatively blurry, you can then pick the ones you want for what your working with, there are tools to do this in mental ray too but it can be easier before hand, perfectly sharp reflections can look very odd and unrealistic.

End result is to get:
A filled in, netural set of high dynamic range images going from super small and blurry to larger and sharper for use in lighting and / or reflections.

The compositing side has more too it than this, these are the general steps needed to get nice, predictagble linier results from a hdr image.

FX supervisor - double negative

Quote