Hello again!
I've recently come across an article where the author mentions FFT3dFilter and shows an example of two pass filtering with it (i.e. applies it 2 times). I doubt that it makes sense, but I'm not sure of it because I don't know inner workings of the filter in detail. So I need to clarify those.
The manual says that image is split into those blocks and then each block's data is separated into a range of frequencies. Then some of the frequencies are thrown away by the filter. The key question is whether the range of frequencies for any block is constant, like in DCT transform (from minimum possible to maximum possible) or it is block-dependent (from min freq present in the block to max present in the block). If I presume that the 4 sigma parameters of the filter define the frequency filter power in 4 subranges of the frequency range for a block, then in the first case applying the filter 2 times is useless, because the frequency range remains the same and the same effect can be achieved by filtering just once. But in the second case this could make sense because after first filter pass the frequency range can change and the second pass will bring different results than 1 pass with different parameters. So what is it?
And another important question - is the frequency range influenced by block size?
Hello again! I've recently come across an article where the author mentions FFT3dFilter and shows an example of two pass filtering with it (i.e. applies it 2 times). I doubt that it makes sense, but I'm not sure of it because I don't know inner workings of the filter in detail. So I need to clarify those. The manual says that image is split into those blocks and then each block's data is separated into a range of frequencies. Then some of the frequencies are thrown away by the filter. The key question is whether the range of frequencies for any block is constant, like in DCT transform (from minimum possible to maximum possible) or it is block-dependent (from min freq present in the block to max present in the block). If I presume that the 4 sigma parameters of the filter define the frequency filter power in 4 subranges of the frequency range for a block, then in the first case applying the filter 2 times is useless, because the frequency range remains the same and the same effect can be achieved by filtering just once. But in the second case this could make sense because after first filter pass the frequency range can change and the second pass will bring different results than 1 pass with different parameters. So what is it? And another important question - is the frequency range influenced by block size?