A novel method for two-dimensional phase estimation

Estimating the phase of a complex radar signal has many important applications in remote sensing. In the one-dimensional (1D) case, the phase estimate is used to dechirp the radar signal in the azimuth direction. In two dimensions (2D), phase estimation can provide elevation data for topographic mapping from radar pairs. Much has been written about 1D phase estimation, and several robust 1D phase estimation algorithms have found wide application. These algorithms, however, cannot readily be extended to two dimensions. If a topographic map is to be generated from a radar pair, then almost invariably a 2D phase unwrapping algorithm is required. Successful unwrapping of two-dimensional phase is sensitive (i) to registration accuracy, subpixel registration accuracy often being necessary, and (ii) to signal-to-noise (S/N) ratio, which cannot be too low. The algorithm presented estimates the phase of a 2D signal by first estimating the phase derivative in two orthogonal directions. The derivative estimator presented introduces no additional phase distortions, and is not particularly sensitive to speckle noise. From the phase derivative the phase is estimated by performing a two-dimensional integration. Simulation experiments have shown that this new method is less sensitive to low S/N ratio than the conventional methods, and that registration errors of as much as one or two pixels do not not affect the accuracy of the phase estimation nearly so drastically as they do in the conventional phase unwrapping algorithms