Dual ISO processing #306
Dual ISO processing #306
Conversation
This update focuses primarily on solving 2 of the biggest dual ISO issues: the flickering issue and artifacts in blown highlights areas affecting lower bit depth footage (e.g. 12-bit, 10-bit…). Added manual control over 3 most crucial parameters for proper exposure blending: horizontal pattern, exposure correction, and black delta. Big performance boost by enabling multithreading. It also includes several minor UI updates/fixes not specific to dual ISO.
bring back spaces for better readability, better precision for black level spinbox, compatibilty between old and new dual iso receipts, bring dual iso help to tooltip, cleanup unused code
was missing for test file
|
Test:
|
|
Export successful now. |
|
Nice! There's a slight speed increase with this new modified parallel AMaZE. Visually, everything looks the same, except I noticed a few bytes smaller DNG file sizes after export. So, I tried to compare the two images with a difference blend mode and amplify the changes, and found a difference, mostly in the middle of the image. 
This is a huge amplification and probably not a big deal. I'm just wondering if there's some overlooked change in the algorithm. For me, it's hard to tell which is the right one or better in terms of image processing. |
|
Thanks for testing! Yes, the difference using AMaZE multithreaded was to be expected and is one of the downsides of this method. Looks your example uses 2 threads. I don't remember my test footage which was showing this also in a standard clip with basic AMaZE debayering. In those days I switched often to RCD - here multithreading was realized in a different way and you don't get those artifacts. Would this also work with DualISO (sry... I did not try to understand to much in detail how exactly DualISO works)? |
2 participants
Using the changes from @fijha and inserting my requests mentioned in the PR.