neat profiling

general questions about Neat Video
Post Reply
jambocan
Posts: 2
Joined: Mon Nov 24, 2008 6:58 am

neat profiling

Post by jambocan »

Great software btw ! :D

In the user guide it suggests using YCbCr profiling for just about everything. What are the specific times you would use RGB and what are the differences in the resulting profiling? What if you were dealing with RGB originated imagery like say a cineon (.cin) file ? I might want to use this for cleaning up a vfx plate that is a scanned image in rgb, but I don't understand how your profiling works to know what color space to use for the profiling based on your user guide. Are these some sort of chebychev or butterworth type of bandpass filters or are you doing something fancy and totally unique? Can you elaborate a little and explain how it fits into your suggestion of always using YCbCr and when you would prefer RGB profiling? Thanks for everything so far!

j
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Post by NVTeam »

Neat Video filter works best when it can separate luminance and chrominance components of the video data, to treat them separately. Using the YCrCb working color space helps doing that, while the RGB color space prevents that. So, unless you want to filter only one (or two) specific channel(s) out of three - R, G, and B, you are advised to use the default YCrCb working color space to let Neat Video work efficiently on the whole set of video data.

Hope this helps,
Vlad
jambocan
Posts: 2
Joined: Mon Nov 24, 2008 6:58 am

neat profiling

Post by jambocan »

Hello Vlad,

Thanks. Yes, I understand that, but could you be a little more specific about how the profiling works ? I can imagine that rgb profiling might work well for instance with film based material, since its noise (grain) is quite unique in each channel (being separate layers of a physical media in rgb), but without knowledge of how your bandpass filters work, I cannot determine this to be true or not. Why do you even have an RGB filter if you don't recommend it in any circumstance? Can you explain a situation where one would use the RGB filtering over the YCrCb, and what things to watch out for when using RGB?

I simply want to understand how the profiling works in order to better use the tool. Temporal filtering is well understood (and explained in your user manual), but you seem elusive about the spatial profiling, critical of other methods (like FFT) without explaining how yours works and why it is better?

j
NVTeam
Posts: 2745
Joined: Thu Sep 01, 2005 4:12 pm
Contact:

Re: neat profiling

Post by NVTeam »

jambocan wrote:I can imagine that rgb profiling might work well for instance with film based material, since its noise (grain) is quite unique in each channel (being separate layers of a physical media in rgb), but without knowledge of how your bandpass filters work, I cannot determine this to be true or not.
Why not, you can directly check that using some sample footage of that special type and process it in different working color spaces. Please try that and check whether RGB produces any visually-better results than the default YCrCb. I expect there will be no improvement in RGB, or YCrCb will perform better. If not then we will all learn something new about use of Neat Video and I, for example, will be glad to add new knowledge to the user guide to help all users.
jambocan wrote:Why do you even have an RGB filter if you don't recommend it in any circumstance?
Because I see it useful only in very special cases. I have mentioned one example of such special case above. Your question about special type of input media offers another example. Neat Video offers a way to use an alternative color space, so you can do such special cases when you know you need them, but in our experience, the default YCrCb working color space has been the best from the standpoint of filtration quality and I haven't yet seen an example of the opposite. That's why I recommend to use the default. But if you will find the RGB working color space is performing better for that special type of input data then it will be very interesting to see a practical example of that and to describe such special case in the documentation. So, please keep this topic updated on any findings regarding use of the RGB space with special type of input data.

Thank you,
Vlad
Post Reply