A nonlinear image representation in wavelet domain using complex signals with single quadrant spectrum

High frequency structures, such as edges, texture, contain significant portion of the information available in images. For coding purposes, the image representation has to provide direct access to this information content without introducing any redundancy. In this paper, we introduce such a nonredundant representation based on complex signals in wavelet domain, whose spectrum has only one nonzero quadrant. The corresponding magnitude and phase responses are linked with the most significant edge properties, namely energy, location and orientation. Simulations show that phase responses provide unbiased estimates of the edge location and orientation. We also discuss our coding strategies for this new representation.