Helicon DNG output and Topaz Photo AI

[clausgiloi]

My workflow consist roughly of:

Sony RAWs -> Helicon DNG -> Topaz DNG -> Lightroom DNG

When I use Adobe DNG converter directly to convert a RAW file to DNG, then Topaz treats this DNG as a "raw" file, but if I stack the raws in Helicon first and output a DNG, then Topaz does NOT recognize this as a "raw" file in the same way and applies different (wrong) corrections.

I'm not saying this is Helicon's fault, but does anyone know why this is happening and if there is anything I can do to fix it? It causes Topaz to change the colors in a way that causes big problems.

Thanks!

[Catherine]

Helicon Focus performs debayering of the image (or, rather, the Adobe DNG toolkit inside Helicon Focus does that). That's about the only thing we can think of. But also, the Adobe code completely creates a new DNG file from scratch when you save it, and who knows exactly what's going on there.

[clausgiloi]

Is it not possible that there are parameters in the creation of the output DNG via the Adobe SDK that could be affecting this?

[Catherine]

No, there are no parameters. And either way this is a valid DNG file, accepted by other various software.

[fdisilvestro]

Hi, not all DNG are equal, you can think of DNG as a container.
When you convert directly a raw file to DNG, the image data typically is not processed and is kept in its mosaiced (RGB) form. Topaz Photo AI treats it as a raw file accordingly, because has AI models for them.

Another type of DNG is the so called "linear DNG" which is a partially processed file, as those generated by Helicon Focus. Here the image has been "demosaiced", (now you have RGB values for each pixel), but it has not been encoded to a color profile yet. Unfortunately, Topaz Ai does not have AI models for many of the linear DNGs, as you have experienced. At this point you would be better off generating a rendered format such as tiff, and input that to Topaz AI.

One option is to use Lightroom to complete the processing of the Linear DNG, and then export to tiff.

[Catherine]

Thank you, that is a good explanation, and a plausible guess as to why Topaz doesn't like our output files.

[clausgiloi]

Hi, not all DNG are equal, you can think of DNG as a container.
When you convert directly a raw file to DNG, the image data typically is not processed and is kept in its mosaiced (RGB) form. Topaz Photo AI treats it as a raw file accordingly, because has AI models for them.

Another type of DNG is the so called "linear DNG" which is a partially processed file, as those generated by Helicon Focus. Here the image has been "demosaiced", (now you have RGB values for each pixel), but it has not been encoded to a color profile yet. Unfortunately, Topaz Ai does not have AI models for many of the linear DNGs, as you have experienced. At this point you would be better off generating a rendered format such as tiff, and input that to Topaz AI.

One option is to use Lightroom to complete the processing of the Linear DNG, and then export to tiff.

Thank you, that is consistent with everything I've learned. It does not seem like there is a fundamental barrier for Topaz to add the functionality to perform their denoise on the de-bayerised data. I will continue to engage with them.

Alternatively, assuming that de-bayerization is reversible, Helicon Focus could include an option to do so? As it is, the promise of RAW to DNG development falls a bit short in my scenario since not all downstream apps apparently can handle the linear DNG (not saying that's Helicon's fault).

If I export a TIFF instead, Topaz certainly won't raw denoise that either, so I'm not sure why I would be better off.

Thanks!