A comparison of spotlight synthetic aperture radar image formation techniques

Space-Variant Post-Filtering for Wavefront Curvature Correction in Polar-Formatted Spotlight-Mode SAR Imagery

“…Furthermore, CBP is not considered a computationally efficient algorithm to implement on a parallel computer, as it typically requires the programmer to make a tradeoff between excessive memory requirements and significant communications overhead on distributed-memory computers [33]. Additionallyj the application of autofocus algorithms for uncompensated platform motion is difficult when using CBP.…”

“…Furthermore, while SAR image formation may be considered a tomographic process, the polar-format algorithm is often unfairly criticized for having unreasonable interpolation errors [71,72], which is not the case when applied to most SAR imagery (the exception being UW-B SAR, as described further in Chapter 6). It is important to realize that for spotlight-mode SAR, whereby the angle of aperture extent 40 is typically small and the Fourier data are offset significantly from baseband, the less computationally burdensome polar-format algorithm can very effectively form images since the polar-to-rectangular interpolation step results in a negligible amount of interpolation error [33]. It has been pointed out in several papers including [71,73], the resampling portion of the polar-format algorithm can be computationally burdensome.…”

A Rationale for Backprojection in Spotlight Synthetic Aperture Radar Image Formation

“…This does not come for free since the two-dimensional interpolation is itself an expensive operation. Still, practice has shown that the combined operation of two-dimensional interpolation followed by inverse twodimensional inverse FFT, direct Fourier inversion, is still faster on most computing architectures than convolutionbackprojection, purpose-built, workstation clusters, or parallelprocessor computers notwithstanding, but this is a complicated subject and results vary [47], [48] and should be considered in light of the aforementioned fast backprojection algorithms.…”

“…Any differences between a convolution-backprojection reconstruction and a direct Fourier inversion reconstruction are due to numerical choices such as interpolation, windowing, assignment of Jacobian values in non-trivial distributions, and use or non-use of portions of the polar data that do not get incorporated in the two-dimensional interpolation of a direct Fourier inversion. Comparison studies are reported in [47], [50], [48], [51], and [52]. Some ruminations on this topic are in [19].…”

Data‐driven sampling pattern design for sparse spotlight SAR imaging

“…According to the projection‐slice theorem, Equation () can be considered as a 2D Fourier transform of the scattering function. When the scene is discretized into cells in both directions and stacking the backscattering coefficients into a vector, we achieve the sparse SAR imaging model in a matrix form $$\begin{equation} \omega =A_\Theta (r)+n, \end{equation}$$where ω is the measured echo data vector; r is the vector of the lexicographically ordered sparse discretized spatial scattering image of the observation area Ω; $$A_\Theta$$ is the Fourier transform based SAR measurement matrix [13]; and n is the additive noise vector of the radar system. The Θ represents the sample location when downsampling is exploited.…”