SAR image processing algorithms based on the ambiguity function

This work focuses on hardware implementations of the ambiguity function defined in terms of signal operators for synthetic aperture radar (SAR) sensor image formation applications. Due to the large quantity of data associated with SAR imaging, basic SAR operations demand high computational cost and are difficult to readily implement on existing commercial digital signal processing (DSP) units. In order to contribute to resolve this problem a search for new efficient algorithm variants in under way compute operations such as discrete Fourier transforms, discrete convolutions, discrete correlations, and Hadamard products of complex signals. These operations are commonly used in the computation of the discrete ambiguity function. Time-frequency tools have been the object of many different studies in SAR image formation operations. The principle behind SAR image formation is recovering a desired reflectivity density function through an inverse convolution process on the raw data generation. It is this particular fact that makes the ambiguity function so important in SAR image formation since it serves to model the impulse response function or point spread function of the SAR system itself. Algorithms for computing the ambiguity designed for this work are being developed in MATLAB and implemented the TMS320C6713 digital signal processor (DSP) unit. The implementation methodology is based on hardware/software algorithm co-design techniques, taking into consideration software and hardware characteristics such as memory cache overflows, twiddle or phase factor pre-computations, extended memory addressing schemes, stride permutation implementations, pipelining schemes, multiply-add DSP architectures, multiple concurrent memory accessing schemes, and overall latency response.