Image Wiener filtering Parallel Realization Based on Wavelet Transform
Sheng Zhong · Signal Processing · 2008
To archieve highly efficient realization of denoising process of large-scale images,a kind of wiener filtering parallel im- plementing method and array architecture based on daubechies 9-7 bi-orthogonal wavelet transform are proposed in this paper.First,a parallel implementing method of lifting wavelet transformation and the PE interconnect are introduced.While the former accelerates the decomposing process,the latter ensures direct connection between PEs which store the same subband coefficients of every transform lev- el.In this way,the communication cost of wavelet transformation and denoising are efficiently reduced.Second,focusing the accumulation problem of subband sliding window during Wiener filtering,the way of cycling prefix-accumulation is adopted to realize the parallel accu- mulation of multi-windows,which dramatically reduces time cost of wiener filtering,and real-time processing requirement of large scale image frames are satisfied.