Title :
Combining the discrete wavelet transform and mixed-domain filtering
Author :
Lazar, Michael S. ; Bruton, Leonard T.
Author_Institution :
Dept. of Electr. & Comput. Eng., Calgary Univ., Alta., Canada
fDate :
7/1/1996 12:00:00 AM
Abstract :
A novel filtering method is proposed that combines the discrete orthogonal wavelet transform (DWT) with the mixed-domain (mixed-D) filtering method. The method uses the DWT to pre- and postprocess those dimensions of the signal that are transformed to the discrete-frequency domain by mixed-D filtering. Using the DWT in this manner provides a controlled mechanism to partition the spectrum of the input signal into subband signals, which then may be selectively filtered during the linear difference equation (LDE) step of the mixed-D algorithm. It is shown that, when the DWT is computed using filters with ideal high- and lowpass frequency responses, the LDE filters used in the mixed-D filtering stage are unchanged by the introduction of the DWT (although the frequency tuple associated with each LDE filter is altered). This indicates that the mixed-D filtering scheme can be easily used in subband coding systems. Results are given for the filtering of a three-dimensional (3-D) linear trajectory signal, representing a common application in video processing
Keywords :
filtering theory; spectral analysis; transform coding; transforms; video coding; wavelet transforms; 3D linear trajectory signal; DWT; discrete orthogonal wavelet transform; frequency tuple; highpass frequency response; input signal spectrum; linear difference equation; lowpass frequency response; mixed domain filtering method; signal dimensions; subband coding systems; subband signals; video processing; Difference equations; Discrete Fourier transforms; Discrete wavelet transforms; Energy resolution; Filtering; Frequency; Nonlinear filters; Partitioning algorithms; Signal processing; Signal resolution; Spatial resolution;
Journal_Title :
Image Processing, IEEE Transactions on