Enfuse 4.1.1 CIECAM bug?

[dkloi]

I've been trying to track down a problem with enblended images suffering a lack of contrast with respect to the remapped images. It seems that I am not the only one experiencing this problem and it is due to the CIECAM blending default option for input images with colour profiles. 

Example: http://www.flickr.com/photos/48606216@N00/10135641463/ The area shown is not in an overlapping region so no blending should occur. The input image looks exactly like the right hand side when CIECAM blending mode is turned off. The left shows a lightening of the shadows with the default options. Enblend 4.1.1 Windows 64-bit version from Sourceforge. Windows 7 64 bit 16GB RAM, i7 3770. Lightroom 4.4 and PWP 6.0.10 64 bit.

Is the problem due to the forward and reverse mappings of the input colour space to CIECAM02 not being inverse? I would have expected that a non-overlapping region should be mapped in the output to exactly the same values as the input but this doesn't seem to be the case with the CIECAM blending mode.

Daniel.

[Bruno Postle]

On 7 October 2013 12:07, dkloi <dkl...@googlemail.com> wrote:

Is the problem due to the forward and reverse mappings of the input colour space to CIECAM02 not being inverse? I would have expected that a non-overlapping region should be mapped in the output to exactly the same values as the input but this doesn't seem to be the case with the CIECAM blending mode.

I don't have an explanation for the difference, but enblend _will_ alter values in non-overlapping areas - The width of the blend is determined by the number of blending levels, so if you want to rule out the blending itself as the source of the problem then you can temporarily set this to -l 1

--
Bruno

[dkloi]

Thanks for the tip about the blending levels. However, even with only a single blending level I get the same lightening of the shadows with the default options. I found a previous post https://groups.google.com/d/msg/hugin-ptx/354rmYe12fA/4cLfTrny11kJ that indicates that the problem could lie with LCMS.

Just noticed 4.1.2 on Sourceforge. Just tried it but it looks like it behaves the same as 4.1.1.

[dex Otaku]

On Monday, 7 October 2013 06:07:39 UTC-5, dkloi wrote:

I've been trying to track down a problem with enblended images suffering a lack of contrast with respect to the remapped images. It seems that I am not the only one experiencing this problem and it is due to the CIECAM blending default option for input images with colour profiles. 

I'm pretty sure this issue started for me with 2013RC1 [on Windows].  I have downloaded RC2 but not processed anything yet.  

I've just gotten used to not being able to render anything successfully without using the --no-ciecam option.

[dex Otaku]

Further info in case it may be of use:

I'm shooting raw with a Sony alpha 55 and a sigma 10-20mm @10mm for panos.  

The camera is set to use the AdobeRGB profile.  I usually use Lightroom [4.x] to make 16-bit TIFFs of the images with some basic exposure processing.  Sometimes I use RawTherapee instead [it has finer control over noise reduction].  My displays are profiled using Dispcalgui/Argyll CMS [under Windows, yes].

Previews using the 16bpc TIFFs display as expected, but no renders expose correctly, regardless of image content.  

Using enblend with --no-ciecam gives "expected" results.  Without it, images are always as described by the OP.

Using the sRGB profile instead of AdobeRGB from the camera makes no difference.  Rendering my intermediates as sRGB [converting colour spaces before Hugin] also makes no difference.

[mh00h]

Does this problem persist in enblend 4.2?

[dkloi]

Not sure whether this particular bug has been fixed, but I have had a problem with coloured speckled pixels in dark areas of the blended output (the speckled pixels weren't present in the remapped input images). An example is attached. This disappears when using the --no-ciecam option. This is with the enblend included in the 64bit Windows version of 2017.0.0.

[mh00h]

I have this exact same speckles problem in Arch Linux. It seems to be worse as images scale up in size. And as you've pointed out, sections of photo that are close to black exhibit the worst of it. It's possible these speckles span into the lighter sections of a photo too but get masked?

I've noticed that most panos I stitch have a flat appearance and am curious if your OP discovery is finally the reason why. I had asked the same question around on reddit but made little progress there.

[dkloi]

I found reference to the bug at

https://bugs.launchpad.net/enblend/+bug/752283

https://bugs.launchpad.net/enblend/+bug/1576862

https://discuss.pixls.us/t/enfuse-artifacts/4687