Estimating the instantaneous frequencies of a multicomponent image by bidimensional empirical mode decomposition

An overview of AM-FM image presentation techniques is presented. Many nonstationary textures are multi-partite, and may be considered to comprise several locally narrowband components. The multicomponent AM-FM model is powerful in describing such images. The empirical mode decomposition is applied to processing textures, and proposed bidimensional empirical mode decomposition (BEMD) followed by a computationally efficient algorithm. It is shown that the components may be demodulated on a spatially local basis based on BEMD. The Burg linear predicting filter is used to reduce the decomposition error of BEMD. A practical approach, the direct method, is proposed for estimating the instantaneous frequencies of each component.