Edge detection using recursive biorthogonal wavelet transform
Michel Barlaud, Thierry Gaidon, Pierre-Philippe Mathieu, Jean-Christophe Feauveau · 1991
A new method of multiresolution edge detection is described. This method is based on a recursive biorthogonal wavelet transform which relates the digital filter to a continuous analysis function, the so-called wavelet. In biorthogonal wavelet analysis, there is a high-pass filter corresponding to the wavelet and a low-pass filter corresponding to the scaling function. The new idea is to compute optimal edge detection filters which yield multiresolution analysis (continuous wavelet). The high-pass filter (infinite impulse filter) is chosen in order to have first derivative properties. The family of low-pass filters is deduced from the high-pass filter using wavelet transform theory relations. The detection filters are proposed as a trade-off between detection, localization and regularity criteria.>