Diversity of edge information using complementary filter banks analysis

We present the complementary filter (CF) banks, a signal processing technique for image processing. The two-dimensional (2D) multirate signal processing theory and complementary filter properties are the base of CF bank design. Procedures to design nonseparable complementary filters for an alias free decimation are presented. The conditions under which perfect reconstruction is achieved with 2-channel bank of filters, using complementary filters are given. Perfect reconstruction is guaranteed when a CF bank is implemented using a quincunx filter and sampling and although no perfect reconstruction for others methods, the analysis and synthesis are performed free of aliasing. We show that CF banks and Wavelet representation are related. The CF banks are implemented with images and their performance is analyzed. The CF banks were applied for edge detection and examples of diversity in edge information obtained for different types of filtering are shown