Directional Filter Banks Forwavelet Decomposition of Images based on the Radon Transform

Directional filter banks offer a useful tool for the decomposition of images in applications where it is necessary to extract features such as edges and textures at different orientations and scales. We introduce an implementation of directional filter banks for the wavelet decomposition of images. The general procedure is similar to the ridgelet transform approach, and it is based on the direct and inverse Radon transforms while relying on angular windowing and ID wavelet transforms in the Radon domain to slice the frequency plane and thus generate the passbands for the 2D filters. The proposed scheme attains increased flexibility in the choices of directions and widths for the passbands, thus allowing different choices of uniform or non-uniform angular sectors in the frequency domain.