Quote:
Originally Posted by Manni
That makes sense, especially if you use medium recovery strength (I use low) and a 200nits target.
I know the grain is in the source (and I don't mind that), I just don't like what MadVr does to it in that instance with the new algo. If you compare the new algo with either old or dumb, you'll see what I mean. It exacerbates the grain due to perceived oversharpening/overprocessing.
|
No, it does not. It's quite the opposite. Tone mapping compresses a very wide dynamic range into a very small range. Which means that the distance between e.g. brighter and darker grain dots in bright image sections becomes compressed, too (both mathematically and perceptionally). Practically that means that grain becomes softened through tone mapping.
The new algo simply reconstructs the original distance (on a perceptional scale) between the grain dots. There's no exacerbation, no oversharpening and no overprocessing going on here at all. You can't use a softened image (old/dumb) as a reference for how grain should be reproduced.
To sum up: The new algo does not exacerbate the grain. The old algo and dumb mode soften it.
Quote:
Originally Posted by magic144
I sent you a PM with links to the logs I just created.
|
Thanks, I'll have a look.