A new class of non-separable symmetric wavelets for image processing

It is known that wavelet analysis is a powerful mathematical tool for image processing. For such type of applications, symmetry of the wavelet filters is claimed to produce less visual artifacts than non-linear phase wavelets. On the other hand, the filters themselves can be separable or non-separable. While separable filters offer the advantage of low-complexity processing, their non-separable counterparts have more degrees of freedom and hence allow better designs. In this talk we discuss about new classes of non-separable wavelet filters with different types of symmetry. A scheme for their construction is given and some applications to edge detection over geometrical images and over industrial data are shown.