Negative pixel values in source footage
Posted: Tue May 28, 2019 3:13 pm
Hi.
I have been handed a few sequences with negative values (in this case EXR converted from Alexa, but problem have been seen before). The values goes down to -0.0005 in one or more channels.
The negative values seems to create some artifacts in the denoise process. Can't show any footage (NDA), but the error looks similar to the spots you get when setting the Noise Levels->Cr/CB values to low.
Is there a "best practice" to get around this? I havent been able to find negative values or plate preperation mentioned in the documentation.
My own suggestions would be one of the following:
1: Clamp negative values to 0 or 0.000001. This is a destructive process, so not happy about that.
2: Offset the footage slightly to all positive values, denoise, then offset footage back again. This might (?) effect the grains as they are being calculated on a plate with other than original values.
Also, is Neat Video supposed to be used before or after any LUT-convertion nodes (e.g. log-C -> lin), or is it just controlled with the "linear input data"-checkbox?
I am working in Nuke using Neat Video v3 and v4 (working at two different companies).
Thanks in advance
I have been handed a few sequences with negative values (in this case EXR converted from Alexa, but problem have been seen before). The values goes down to -0.0005 in one or more channels.
The negative values seems to create some artifacts in the denoise process. Can't show any footage (NDA), but the error looks similar to the spots you get when setting the Noise Levels->Cr/CB values to low.
Is there a "best practice" to get around this? I havent been able to find negative values or plate preperation mentioned in the documentation.
My own suggestions would be one of the following:
1: Clamp negative values to 0 or 0.000001. This is a destructive process, so not happy about that.
2: Offset the footage slightly to all positive values, denoise, then offset footage back again. This might (?) effect the grains as they are being calculated on a plate with other than original values.
Also, is Neat Video supposed to be used before or after any LUT-convertion nodes (e.g. log-C -> lin), or is it just controlled with the "linear input data"-checkbox?
I am working in Nuke using Neat Video v3 and v4 (working at two different companies).
Thanks in advance