Deband
- JPulowski
- Topic Author
Please Log in or Create an account to join the conversation.
- lowenz
C:\ReShade 4\Shaders\Deband2.fx(136,2-13): error X3511: unable to unroll loop, loop does not appear to terminate in a timely manner (212 iterations) or unrolled loop is too large, use the [unroll(n)] attribute to force an exact higher number
Please Log in or Create an account to join the conversation.
- lowenz
Please Log in or Create an account to join the conversation.
- JPulowski
- Topic Author
Seems due to changing tex2Dlod to tex2D. Thanks, will fix it in the next beta.lowenz wrote: C:\ReShade 4\Shaders\Deband2.fx(86,15-31): warning X3553: can't use gradient instructions in loops with break, forcing loop to unroll
C:\ReShade 4\Shaders\Deband2.fx(136,2-13): error X3511: unable to unroll loop, loop does not appear to terminate in a timely manner (212 iterations) or unrolled loop is too large, use the [unroll(n)] attribute to force an exact higher number
Please Log in or Create an account to join the conversation.
- Enterprise24
Please Log in or Create an account to join the conversation.
- JPulowski
- Topic Author
Changelog:
- Added preset based basic setting and moved thresholds to advanced settings
- Fixed the problem where for loops were unable to be unroll in d3d9
- Added gradient angle checking (disabled by default)
I think the new thing is the gradient angle checking. From my limited testing, I did not see much benefit from it. It also causes a decrease in performance. Depending on your feedback I might remove it all together. It might be more efficient in video content and current parameters might not be optimized for video game content. Basically, when it detects a significant a gradient angle it boosts the thresholds for further debanding. Feel free to play with in addition to standard thresholds. (it is controlled by maximum angle and angle boost parameters)
I have also added madVR-like preset system. The current preset parameters are based on madshi's notes and do not reflect the final parameters. That actually depends on your feedback. So feel free to test them and tell me what do you think. You can also use "Custom" preset to input your own values in the advanced section. The default values in the advanced section are based on the medium preset by the way.
Again, if you encounter any bugs, errors or warnings feel free to let me know.
Please Log in or Create an account to join the conversation.
- lowenz
Please Log in or Create an account to join the conversation.
- Enterprise24
Old deband have more strength but there is noticeably flicker from noise.
New deband cause much less flicker (on low setting - don't wan't to touch medium or higher since flicker is more noticeable).
Please Log in or Create an account to join the conversation.
- Scorpio82CO
Please Log in or Create an account to join the conversation.
- JBeckman
Scorpio82CO wrote: Doesn´t work for me in reshade 4.2.1, could you update the version??
Do you have the ReShadeUI.fxh file?
From this new version of the shader.#include "ReShadeUI.fxh"
github.com/crosire/reshade-shaders/blob/...haders/ReShadeUI.fxh
raw.githubusercontent.com/crosire/reshad...haders/ReShadeUI.fxh
That's accompanying ReShade.fxh now since a few changes back to the shader files but since Deband hasn't updated in some time it might still have been from before that change got added to the shader repository.
Unless ReShade.exe has a way to check for changes or you periodically re-download everything or manually check the shader Github repository it's very easy to miss changes to shader files even for core functionality such as this adjustment.
Just a suggestion, logger in the ReShade UI itself might have some additional info from breaking errors (red color.) to warnings (yellow color.) and perhaps more.
Please Log in or Create an account to join the conversation.
- Duran.te
Since pictures are worth a thousand words, here's some screenshots from Sekiro - Colours Band Twice:
(Yep, once again with over saturation)
Values are similar to default ones, with just few changes:
- The Old Debanding Initial Radius has been changed from 16.000 to 24.000 like beta versions.
- Threshold from Beta Debandings modified to 1.02
(in the screenshot, I used the default "Low" strength preset with the updated version to show the new preset system behaviour)
- Iterations of every version edited to 8 (2x max value) to better highlight the differences.
[hr]Huge fan of the new menu setting: The preset system is very intuitive and the Advanced line it's a definitely cleaner solution!
I've also tested the new gradient angle check trying to find out if there were significant difference with it on (except for performance decreasing
Please Log in or Create an account to join the conversation.
- JPulowski
- Topic Author
Here is the next beta, which is also the release candidate.
Changelog:
- Removed gradient angle checking since it doesn't help with debanding, in the future i will probably replace it with a better suited local spatial frequency activity analysis function (e.g. contrast, uniformity, entropy etc.).
- Cleaned up the code.
- Added a new "Debug view" parameter which should help with determining range and iterations to make sure that the blurred image captures all the banding in the image.
Please Log in or Create an account to join the conversation.
- Duran.te
An advice I could give, according to my experience, would be to boost up a little bit the dithering effect as, in many cases, just raising up each threshold is not enough to get rid of color banding without loosing too many details.
Also, it could be added an extra very high profile for all of those gradient-based games with excessive banding issue.
These are some few examples:
From top to bottom: Inside (2016) - Ibb & Obb (2014) - Necropolis (2016)
Please Log in or Create an account to join the conversation.
- empleat
I tried latest bundle, i found: reshade 1.1 + sweetfx 2.0 and it didn't worked, while default shaders did.
I highly increased treshold to 0.12 and nothing.
I added in - i don't know exactly how it called, there were values like this: USE_CARTOON 0.
So i simply wrote USE_DEBAND 1 - that's how i named my shader and i pressed scroll lock and it said it is on so.
No idea why it didn't worked.
Please Log in or Create an account to join the conversation.
- lowenz
Please Log in or Create an account to join the conversation.
- empleat
- lowenz
CTRL+click and you'll get the DLLs (ReShade32.dll + ReShade64.dll)empleat wrote: ? There is no dll file...
Please Log in or Create an account to join the conversation.
- JPulowski
- Topic Author
Thanks for your feedback. It doesn't have anything to do with dithering. Dithering is already optimized for 8 bpc output. The problem is current implementation is static which means for every region the shader applies the same thresholds. Which is why you either have too much debanding or it's good but still some banding remains. In order to overcome that I need to apply a spatial frequency analysis function to moderate thresholds. Current version only has additional sanity checking compared to the previous ones, but the thresholds themselves are still static, they are not adapting to the image content (gradient vs. texture). I am already reviewing some known methods and will push an update once I have the time to develop a prototype, again I will share it here first, then will make it public. Also feel free to suggest values for the very high preset.Duran.te wrote: These values are pretty okayish.
An advice I could give, according to my experience, would be to boost up a little bit the dithering effect as, in many cases, just raising up each threshold is not enough to get rid of color banding without loosing too many details.
Also, it could be added an extra very high profile for all of those gradient-based games with excessive banding issue.
Please Log in or Create an account to join the conversation.
- empleat
Because it is more simple and i put into mpc folder ReShade32.dll and Deband.h file and which i renamed to Reshade.fx. Still not working.
I used reshade 2.0 there is a link at first post on site 1.
Please Log in or Create an account to join the conversation.
- JPulowski
- Topic Author
Please Log in or Create an account to join the conversation.