Evaluation of angular interpolation kernels in fast back-projection SAR processing

An implementation of a fast, factorised, back-projection scheme is presented which is based on interpolation of range-angle sub-aperture data. A method for determining the required angular sampling rate is proposed and the implementation scheme is evaluated for nearest neighbour and cubic interpolation kernels against image quality and number of operations required. A SAR simulator is used for investigation of the performance and to trade the algorithm parameters. The impulse response function, using interpolation kernels at different sampling rates, is evaluated in terms of peak side-lobe levels and compared with a global back projection implementation. The results show that good processing performance is obtained even when high image quality is required when using cubic interpolation and an adequate sampling rate. This approach is thus suitable for challenging microwave SAR processing problems.