Comparative study of scalar and multiwavelet filters in transform-based compression of IRLS images

Multiwavelets are a new addition to the wavelet theory of discrete numerical transform. Realizable as matrix-valued filter-banks leading to wavelet transform bases, multiwavelets offer simultaneous orthogonality, symmetry and short support, which is not possible with scalar 2-channel wavelet filter-banks. We study two DmWT (Discrete multiWavelet Transform) decomposition (RR "repeated row" and PF "critically sampled preprocessing") algorithms versus the scalar dyadic wavelet transform. We compare the compression efficiency of these algorithms by retaining a given number of largest coefficients in the 2-dimensional transform of test images and measuring the MSE approximation error. In systematic experiments we change the IRLS (infrared line-scan) test images, decomposition algorithms as well as the filters type/order. The obtained computer results prove the significant improvement of the DmWT over the scalar wavelet transform for comparative filter order and moderate higher computational complexity.