My two requests for new features
Posted: Sun Jun 10, 2018 9:12 pm
I would have thought that others would have requested these already, but I am not seeing them as I look through the forum questions and comments.
Two features that I think would help NI considerably are
1) non-rectangular and/or multiple noise model areas and
2) something to address periodic noise, such as regular lines across an image or halftoning
Non-rectangular or multiple noise model areas:
Very often, it is difficult to find a single featureless rectangular area for NI to use as a noise model. Why can't the noise model be selected using an irregular boundary in a manner similar to the lasso tool in Photoshop? Often I have plenty of area in my image that is representative of the noise I wish to suppress, but no sizeable, contiguous rectangular portions. As a result, I frequently have to choose very small areas as my noise model and the small areas aren't necessarily respresentative of the noise throughout the whole image.
Alternatively, would it be possible to use multiple rectangular selections as a representative noise model? While not as convenient from the user perspective as an irregular boundary, at least multiple noise model areas would allow much better noise modeling when an image has few featureless areas to work with.
Periodic noise:
Certain noise is periodic and Fourier, wavelet and other methods can be very powerful in suppressing that sort of noise. I am truly impressed by NI's capability to suppress most forms of noise, but when it comes to periodic noise, NI is weak and it appears that frequency transform based methods are not used at all within the program. Halftoning can only be suppressed by using fairly low frequencies which lead to more blurring of the base image than is necessary. Given the simplicity of Fourier and similar methods (wavelets would be better because the suppression can be restricted to portions of the image as well as frequencies), I am a bit surprised by their omission. NI appears to base much of its work on image gradients, an approach that I strongly support, but adding traditional frequency-based methods would be a big help when periodic noise is encountered.
Thanks for your consideration
Bruce G
Two features that I think would help NI considerably are
1) non-rectangular and/or multiple noise model areas and
2) something to address periodic noise, such as regular lines across an image or halftoning
Non-rectangular or multiple noise model areas:
Very often, it is difficult to find a single featureless rectangular area for NI to use as a noise model. Why can't the noise model be selected using an irregular boundary in a manner similar to the lasso tool in Photoshop? Often I have plenty of area in my image that is representative of the noise I wish to suppress, but no sizeable, contiguous rectangular portions. As a result, I frequently have to choose very small areas as my noise model and the small areas aren't necessarily respresentative of the noise throughout the whole image.
Alternatively, would it be possible to use multiple rectangular selections as a representative noise model? While not as convenient from the user perspective as an irregular boundary, at least multiple noise model areas would allow much better noise modeling when an image has few featureless areas to work with.
Periodic noise:
Certain noise is periodic and Fourier, wavelet and other methods can be very powerful in suppressing that sort of noise. I am truly impressed by NI's capability to suppress most forms of noise, but when it comes to periodic noise, NI is weak and it appears that frequency transform based methods are not used at all within the program. Halftoning can only be suppressed by using fairly low frequencies which lead to more blurring of the base image than is necessary. Given the simplicity of Fourier and similar methods (wavelets would be better because the suppression can be restricted to portions of the image as well as frequencies), I am a bit surprised by their omission. NI appears to base much of its work on image gradients, an approach that I strongly support, but adding traditional frequency-based methods would be a big help when periodic noise is encountered.
Thanks for your consideration
Bruce G