wavelet transform problem
in [Matlab]
|
From: Fung Lee on 11 May 2010 08:24 I want to know why there is a coefficient expansion when using matlab wavelet toolbox to perform wavelet decomposition. For example, using function dwt to do 1 decomposition to an image with size of 512x512 with 'bior5.5' filter but in each subband the size is 261x261 instead of 256x256. I get really confused with that. I hope someone can help me with that. thank you in advance.
From: Wayne King on 11 May 2010 08:40 "Fung Lee" <sophiahk111(a)hotmail.com> wrote in message <hsbi94$fgk$1(a)fred.mathworks.com>... > I want to know why there is a coefficient expansion when using matlab wavelet toolbox to perform wavelet decomposition. For example, using function dwt to do 1 decomposition to an image with size of 512x512 with 'bior5.5' filter but in each subband the size is 261x261 instead of 256x256. I get really confused with that. I hope someone can help me with that. thank you in advance. Hi Fung Lee, that is because the default extension mode for the wavelet transform is half-point symmetrization. Try this: dwtmode('per','nodisp'); Then call dwt2() Hope that helps, Wayne
From: Fung Lee on 11 May 2010 11:27 "Wayne King" <wmkingty(a)gmail.com> wrote in message <hsbj7k$fjo$1(a)fred.mathworks.com>... > "Fung Lee" <sophiahk111(a)hotmail.com> wrote in message <hsbi94$fgk$1(a)fred.mathworks.com>... > > I want to know why there is a coefficient expansion when using matlab wavelet toolbox to perform wavelet decomposition. For example, using function dwt to do 1 decomposition to an image with size of 512x512 with 'bior5.5' filter but in each subband the size is 261x261 instead of 256x256. I get really confused with that. I hope someone can help me with that. thank you in advance. > > Hi Fung Lee, that is because the default extension mode for the wavelet transform is half-point symmetrization. Try this: > > dwtmode('per','nodisp'); > > Then call dwt2() > > Hope that helps, > Wayne Thank you for reply. I have just tried in Matlab. It seems that the size is ok as expected, half the dimension. But there is a boundary problem in the decomposed image. I also want to know to know how to have a perfect reconstruction of DWT with symmetric extension. Because I have tried simulating with reconstruction with symmetric extension, but I could not get a perfect reconstruction back. One more question. Biorthogonal filters are used in wavelet application but it is known to be not energy perserving. Is it mean that a perfect reconstruction is impossible if using biorthogonal filters?
From: Wayne King on 12 May 2010 06:52 "Fung Lee" <sophiahk111(a)hotmail.com> wrote in message <hsbt09$nhc$1(a)fred.mathworks.com>... > "Wayne King" <wmkingty(a)gmail.com> wrote in message <hsbj7k$fjo$1(a)fred.mathworks.com>... > > "Fung Lee" <sophiahk111(a)hotmail.com> wrote in message <hsbi94$fgk$1(a)fred.mathworks.com>... > > > I want to know why there is a coefficient expansion when using matlab wavelet toolbox to perform wavelet decomposition. For example, using function dwt to do 1 decomposition to an image with size of 512x512 with 'bior5.5' filter but in each subband the size is 261x261 instead of 256x256. I get really confused with that. I hope someone can help me with that. thank you in advance. > > > > Hi Fung Lee, that is because the default extension mode for the wavelet transform is half-point symmetrization. Try this: > > > > dwtmode('per','nodisp'); > > > > Then call dwt2() > > > > Hope that helps, > > Wayne > > Thank you for reply. I have just tried in Matlab. It seems that the size is ok as expected, half the dimension. But there is a boundary problem in the decomposed image. I also want to know to know how to have a perfect reconstruction of DWT with symmetric extension. Because I have tried simulating with reconstruction with symmetric extension, but I could not get a perfect reconstruction back. > > One more question. Biorthogonal filters are used in wavelet application but it is known to be not energy perserving. Is it mean that a perfect reconstruction is impossible if using biorthogonal filters? Hi, You can have perfect reconstruction with biorthogonal filters. In fact, with biorthogonal filter banks you can get perfect reconstruction with linear-phase filters, something that is not possible with orthogonal filter banks (except for the Haar). Wayne
|
Pages: 1 Prev: wavelet transform problem Next: out of memory while using roots |