Subaperture autofocus for synthetic aperture radar

A subaperture autofocus algorithm for synthetic aperture radar (SAR) partitions range-compressed phase-history data collected over a full aperture into equal-width subapertures. Application of a one-dimensional Fourier transform to each range bin converts each subaperture data set into a full-scene image (map). Any linear phase difference, or phase ramp, between a pair of subapertures expresses itself as cross-range drift in their maps. A traditional autofocus algorithm fits a polynomial to inferred equal-width phase ramps. If the true phase error function contains significant high-order components, then polynomial regression generates a poor estimate of the phase error function. Instead of filling a polynomial, we fit a sinusoidal function through the inferred phase ramps. An example with a degraded SAR image shows how a sinusoidal correction improves image quality. We compare lower bounds on mean squared error (MSE) for polynomial and sinusoidal parameterizations. Sinusoidal parameterization reduces MSE significantly for model orders greater than five