Dear all,
I would like to implement a Koren-Laine wavelet into my project, but I am a little bit confused of one thing. The proposed FIR filter bank looks like this:
g(-1,1)=[2 -2]
h(-2,-1,0,1)=[0.125,0.375,0.375,0.125]
k(-2,-1,0,1,2,3)=[-0.0078125 -0.0546875 -0.171875 0.171875 0.0546875 0.0078125]
My question is - is it possible to implement this filter bank as:
g=[2 -2 0 0 0 0]
h=[0.125 0.375 0.375 0.125 0 0]
k=[-0.0078125 -0.0546875 -0.171875 0.171875 0.0546875 0.0078125]?
Or do I have to shift g to look like this:
g=[0 2 -2 0 0 0].
What could g(-1,1) and h(-2,-1,0,1) mean anyway? Because adding zeros in front of filter causes only a change in delay. And this shift needs to be compensated in convolution to keep detail coefficients on the right place and achieve perfect reconstruction. Isn`t this right?
Anyone any idea?
Thanks for any advice.
I would like to implement a Koren-Laine wavelet into my project, but I am a little bit confused of one thing. The proposed FIR filter bank looks like this:
g(-1,1)=[2 -2]
h(-2,-1,0,1)=[0.125,0.375,0.375,0.125]
k(-2,-1,0,1,2,3)=[-0.0078125 -0.0546875 -0.171875 0.171875 0.0546875 0.0078125]
My question is - is it possible to implement this filter bank as:
g=[2 -2 0 0 0 0]
h=[0.125 0.375 0.375 0.125 0 0]
k=[-0.0078125 -0.0546875 -0.171875 0.171875 0.0546875 0.0078125]?
Or do I have to shift g to look like this:
g=[0 2 -2 0 0 0].
What could g(-1,1) and h(-2,-1,0,1) mean anyway? Because adding zeros in front of filter causes only a change in delay. And this shift needs to be compensated in convolution to keep detail coefficients on the right place and achieve perfect reconstruction. Isn`t this right?
Anyone any idea?
Thanks for any advice.
