The following workflows yield different results. I’m working with some old Canon CRW files here, and that may or may not be part of the issue. Haven’t tried this with newer CR2/CR3 files yet.
- CRW → DxO Pure Raw → DNG → Topaz
- CRW → DxO Pure Raw → TIFF → Topaz
- CRW → Topaz
I am primarily using Photo AI to upscale the images.
The second workflow produces fantastic results.
The first workflow (passing DNG into Topaz) results in odd artifacts such as colors bleeding into small details and color balance changes.
Finally, the 3rd workflow has no issues, but even with sharpening in Photo AI, the results are usually a little better with the 2nd workflow.
This isn’t necessarily a problem, because I just don’t use the first workflow. However I’m curious as to what’s going on here. Does Photo AI not realize that lens corrections have been applied to the DNG file…is that the problem? What’s causing the 2nd workflow to have issues?
This is some great information, and I would like to share some of these insights with the developers in order to get a better understanding as to why certain workflows are working and others are not.
Would you mind sharing some screenshot examples of the edits you are receiving so we can review these workflows?
Okay here are 5 crops from different processing workflows.
To my mind, the best result hands down comes from upscaling a TIFF output from DxO Pure Raw in Topaz with no other Topaz enhancements.
Note 1: sending DxO DNG output to Topaz produces a TIFF output file with raw data – opening that in PS brings up the Camera Raw dialog with custom white balance settings of Temperature +5, Tint +1. I have not played with the white balance controls to see if the color balance could be restored.
Note 2: Sending CRW to Topaz produces a TIFF output file which does not contain raw data. So DNG → Topaz produces raw TIFF and CRW → Topaz does not produce raw TIFF. That seems inconsistent to me. One raw format produces raw output and another raw format doesn’t. This isn’t an issue for me, it’s just something I noticed, and there may be a good reason for it.
Original image was 3072x2048 before upscaling. Sample images below are (roughly) 1650 x 1080 crops from the 2x upscaled image. These are JPEG files with high quality, and I verified that the JPEG compression level did not substantially alter the images.
Images in order of appearance. Topaz control settings are shown in parentheses (in order of appearance in the dialogs).
- CRW image upscaled 2x in Photoshop (no other adjustments)
- Output of DxO Pure Raw upscaled 2x in Photoshop (choice of DNG or TIFF output from DxO makes no difference)
- CRW image upscaled 2x in Topaz with Raw Denoise (6/3), Sharpen (Standard/31/1) and Upscale (2x/hifi-v2/1/1/1)
- DxO DNG output upscaled 2x in Topaz with Upscale (2x/hifi-v2/1/1/1) only
- DxO TIFF output upscaled 2x in Topaz with Upscale (2x/hifi-v2/1/1/1) only
In workflow (3), the sharpening strength (31) was as much as could be applied without generating objectionable artifacts. I also tried other sharpening algorithms (Strong/Lens and Motion Blur), but the Standard algorithm worked best.
Some things to pay attention to in the images.
Overall color balance
Tree line in background
Clarity of grooves in front tire.
Color of front landing gear chrome strut
Chromatic aberration on front edge of horizontal stabilizer (at top of tail)
Elliptical sticker on right-side blade of propeller (compared to workflow 1)
Workflow 1
Workflow 2
Workflow 3
Workflow 4
Workflow 5
Thanks for sending those in to us here with such great detail! These are some great examples for the different workflows and the outputs they create. The differences in the images for each workflow are quite noticeable in the several points you brought up. I agree that the workflow 5 example produces the most desirable edits.
I will be sure to share this with some my team for us to review.
Thanks, John.
Please update me if your team figures out why the results with the DxO processed DNG file are different. I am very curious about that.
1 Like
