AM-FM image analysis techniques

We present an overview of AM-FM image analysis techniques. Many nonstationary textured images are multi-partite, and may be considered to comprise several locally narrowband components. We introduce a multi-component AM-FM model useful in describing such images, and show that the components may be demodulated on a spatially local basis using multiband filtering followed by a computationally efficient nonlinear algorithm. We outline practical approaches both for estimating the locally dominant image frequencies and for computing a full AM-FM image representation. Finally, we discuss how the estimated frequencies may be used to compute flow lines.