Comparison of Luxrender Linear Tonemapper to REAL Camera

[turquoiserabbit]

Hello all. As an experiment, I have a friend with a professional camera and I thought it would be neat to recreate a scene with Luxrender and compare camera settings to Luxrender settings.

I had my friend take a series of 7 pictures from a tripod at different f-stops and shutter speeds. I then modeled a replica of the scene in Blender and rendered out with the same camera settings on Luxrender.

The camera model is a Canon EOS 5D Mark III. Settings that stayed the same were- ISO: 100, Focal Length: 42mm, Exposure program: Aperture priority, No flash.

Essentially I tried to model fairly close to the scene in Blender, then used a sun and hemi with an hdri map as close to the lighting at the time of day we took the shots. I modeled the basic interior of the room including window openings. Then I adjusted the light strengths/gain until the first render matched as close as I could get it to the first photograph. My theory being that if the first ones matched, then if only the camera settings changed that they would all match.

Here are the results (right click - view image to see the whole thing):


The camera settings listed are the same for the images directly above and below each other. As you can see, there is almost no difference in the luxrender images, but a noticeable difference in the camera pictures. The only way I noticed a difference with the lux images was laying them right on top of another and selectively hiding and un-hiding the top image. Otherwise I can't even tell.

Anyway, thought I'd post the results of this silly experiment here. If anyone has any insights as to the reason for the difference between the camera and Luxrender, I'd love to hear them.

[luxRoot]

Looks good, I'd make a custom HDRI of the room as well, but I know it's time consuming.

[Abel]

If anyone has any insights as to the reason for the difference between the camera and Luxrender, I'd love to hear them.

Based on the stop info you provided one would expect all images to be the same, except for the DOF on the images taken by the camera (and on LuxRender as well if it would have been switched on). So, the fact that they look the same in LuxRender is to be expected. As to the differences in the camera, one variable that is hard to control is the lighting so that would be my first suspect for causing differences. To check this, one could create two series of images right after each other and see if the photographs with the same settings match exactly. Or alternatively, you could use a studio setting where you can be sure that the lighting setup will not change.

In any case, nice experiment!

[patricks]

i guess that there is a mistake in the Lux scene light setup .... you can not use HDRI for interior light simulation (artificial light ) ... HDRI based light uses no light falloff. i know a lot of people uses HDRI also for studio light setup or other artificial light setups ... but this in theory is wrong (because of no light falloff) .... HDRI should be only used for outdoor scenes as here the light is much more closer to being without falloff.
at least this is how i understand it ....

Greetings Patrick

[jeanphi]

Hi,

It is perfectly valid to use HDRI lighting for interiors. What you should ensure for good lighting when using HDRI is that the subject stays small compared to the environment simulated by the HDRI. That falloff stuff is just a concept introduced by non physical renderers to try and match what real lights do, the falloff is naturally done in a physically based renderer based on the size of the objects, their distances, ...

Jeanphi

[patricks]

Hi jeanphi

this was not a critic against Lux render ... what i mean is that light has a inverse square light falloff ... but HDRI normally dont match this falloff . exterior light on the other hand has in "practise" no falloff as the sun is so far away that the falloff is not noticeable at all. You can make a simple test to match a scene using HDRI light and then the same scene with mesh lights ... you will see that even you can get the overall light look similar, you will notice that the highlights on the objects are much more dim on the HDRI version then when using mesh lights ... and this is because of the falloff.
of cause i could be wrong and Lux uses inverse square falloff for HDRI environment light too ?

this was more a general thought about the usage of HDRI for any render engine ....

i guess that if turquoiserabbit uses mesh lights instead of the HDRI, Lux would match the photos much closer.

Greetings Patrick

[patricks]

ops ... i just read the first post again and it looks like the light only comes from the exterior through the windows .... so in this case the HDRI should work.

anyway i am still wondering if using a HDRI environment for studio setup is "correct" and the same then using mesh lights (i am talking in general regardless of the render engine).

Greetings Patrick

[turquoiserabbit]

Based on the stop info you provided one would expect all images to be the same, except for the DOF on the images taken by the camera (and on LuxRender as well if it would have been switched on).

I'm glad to know I didn't mess up with my Luxrender settings. DOF was turned on for Luxrender, and in fact my original idea was to compare how the depth of field looked compared to a real camera, but I think perhaps I failed to convey that goal to my photographer friend when he took the pictures, since they are all very much in focus.

As for the lighting being different in the photographs, I can certainly imagine that the position of people in the room (my friend, myself, and a few other members of my family at the time) would cast shadows and change brightness. As well, it's possible that cloud cover changed light levels between shots.

My photographer friend is on holiday for a week, but when he gets back I'll ask him to take some pictures with intentionally strong depth of field blur, and try comparing those.

[SATtva]

Also maybe ask him to use a timer or a remote trigger to make shots with nobody in the room.

[claunia]

Interesting test.

Considering the variance in each photo, all should be equal.

The difference in lighting in the camera is clearly not related to the aperture or exposure, I would bet a cloud blocked some of the sunlight in the darker ones.