Prominent point processing (PPP) is a very popular synthetic aperture radar (SAR) autofocus algorithm, which can estimate errors caused by any type of unknown motion or other error source and has high accuracy under the assumption that a prominent point scatterer is available within an imaged scene. Proposed is an advanced PPP algorithm with a preprocessing technique based on the short-time fractional Fourier transform (STFRFT) to relax the limitation on the scene content. The performance of the proposed autofocus algorithm is validated using real radar data.Introduction: Autofocusing is one of the key steps to gather wellfocused images for airborne synthetic aperture radar (SAR) [1]. Many SAR autofocus algorithms have been developed and used successfully. There exist three basic types of autofocus algorithm [2]: subaperturebased autofocus (SBA) (e.g. map drift), optimisation-based autofocus (OBA) (e.g. contrast optimisation) and point-based autofocus (PBA). The SBA and OBA techniques are mostly suitable for images with low-order phase errors. The well-known PBA technique is prominent point processing (PPP) [1]. This method operates on the received signal before azimuth compression and provides direct measurement of the phase history from a prominent point scatterer. The PPP algorithm has been proved to be very accurate and reliable, because we do not assume any specific model of the phase error function. However, the PPP needs a prominent point to obtain properly focused images. Moreover, if the signal from a prominent point is not dominant relative to the summation of same-range targets, the phase history will not provide an accurate reference phase (see Figs. 1a and b). This Letter presents an improved PPP approach with a preprocessing technique based on the short-time fractional Fourier transform (STFRFT). Unlike the original PPP method which needs a very high signal-to-background ratio of the prominent point, the proposed autofocus approach is an efficient improvement, which can enhance the ratio by filtering in the short-time fractional Fourier domain (STFRFD), and it becomes robust against limitation on the scene content.