Image compression using fractal multiwavelet transform (Q2004391): Difference between revisions

In general, a single wavelet constructed from a multiresolution analysis (MRA) using only one scaling function could not satisfy orthogonality, symmetry and compact support simultaneously except for Haar wavelet. By contrast, multiwavelets from MRA using multi-scaling functions could do. This paper mainly reviews MRA using multi-scaling functions, multiwavelet functions, multi-scaling filters and multiwavelet filters first, then defines discrete multiwavelet transform in matrix form by using multi-scaling filters and multiwavelet filters, where the length and support of wavelet multifilter could be equal or larger than the length and support of scaling multifilter, for finite-length signals or image applications. Similarly, fractal multiwavelet transform in matrix form is also defined for MRA using affine fractal multi-scaling functions. Finally, discrete multiwavelet transform with Chui-Lian multifilters using orthonormal multi-scaling functions and multiwavelet functions on [0, 2], and fractal multiwavelet transform using the well-known DGHM scaling functions are primarily compared with traditional Haar and D4 wavelet transforms for image compression application by resulting entropy and PSNR values. Notice that the biorthogonal wavelet transform with four vanishing moments of both its wavelet and dual wavelet is popular in image compression, if you are interested in, you could compare these multiwavelet transforms with it further using more images.