Odd behavior when compositing OpenEXR

[nate101]

Hi All,

Apologies in advance if this is a noobish question. I'm using IM 7.0.5-Q16-HDRI for Windows for whatever it's worth. I'm trying to composite two EXR files using the linear dodge method but so far have been unable to replicate the output I get from Photoshop. I've added links to the following items for reference:

• A. "Shading" base layer EXR LINK

• B. "Reflection" layer EXR (to be layered on top w/ linear dodge) LINK

• C. "Composite Reference" EXR (how A + B should look when properly composited) LINK

• D. Photoshop file for reference showing proper compositing behavior (Shading + Reflection = Composite Reference) LINK

I've tried to composite A and B in IM using the following command:

Code: Select all

magick Sphere-Shading.exr Sphere-Reflection.exr -compose linear-dodge -composite Sphere-IMComposite.exr

You'll see if you compare the IM output from this command with item C above (the reference exr) that there are noticeable differences around the edge of the sphere and also in the sphere's specular highlight. I've tried a bunch of things but still haven't been able to figure out what I'm doing wrong. Not sure if its a syntax issue or some setting in IM I'm missing.

Appreciate any help anyone can give.

[fmw42]

This command works fine for me in IM 7.0.7.11 Q16 HDRI Mac OSX. I get the same result as your reference EXR image.

Code: Select all

magick Sphere-Shading.exr Sphere-Reflection.exr -compose linear-dodge -composite Sphere-IMComposite.exr

But if I save to PNG, I get a darker green. PNG does not keep the wide range of color values as EXR does. EXR can have values outside the 16-bit range that only get saved properly using a format that allows such values in HDRI, such as EXR, HDRI, PFM and MPC (and TIFF with the right define).

[nate101]

Thanks Fred maybe just a boneheaded mistake on my part. Will retest and see if I can duplicate.

[fmw42]

It is possible EXR was fixed after IM 7.0.5. So if it fails for you, try upgrading IM.

I checked 7.0.5.4 and still get the correct result.

P.S. I am viewing using Mac Preview and all seems fine colorwise. There is a slight difference in the antialiasing at the border of the sphere and background, which may be the cause of the 7% difference I get when doing

Code: Select all

magick compare -metric rmse Sphere-CompositeReference.exr Sphere-IMComposite_im705.exr null:

4682.02 (0.071443)

I also see slight color difference between your reference exr and the one from Photoshop.

[nate101]

Yes the anti-aliasing around the edge of the sphere was one of the things bugging me. PS doesn't seem to have this antialiasing issue from what I could tell. Even more concerning, I was seeing a difference in the white highlight on the sphere (not sure if you saw that too), which made me wonder if the brightness values in those pixels were being clamped somewhere along the line.

Going to upgrade to 7.0.7 and retest and will let you know if I notice any difference.

[fmw42]

I see little color change if at all between my IM version and your Reference. There is an antialiasing difference. But I see more color change between the PS versions and your Reference, but less aliasing difference.

There is no difference between my using IM 7.0.5.4 and 7.0.7.11.

Code: Select all

magick compare -metric rmse Sphere-IMComposite_im705.exr Sphere-IMComposite_im70711.exr null:

0 (0)