Video-SAR systems that employ backprojection to generate frames offer efficient re-use of processed data in the image domain, minimal distortion due to wavefront curvature, a fixed frame viewing angle, and several other advantages over Doppler-based methods such as polar formatting. However, one challenge that persists is real-time autofocusing SAR imagery at high frame rates on small airborne radars with limited computational capacity. The work herein describes a method to condition the I & Q data before backprojecting onto the image grid so that blurring due to errors on the motion measurement system is minimized. The method takes advantage of the extremely efficient FFTs on Graphical Processing Units (GPU) to transform a select number of ranges lines to the Doppler domain so that phase gradient autofocus can be performed. The subsequent autofocus vector is fedback to the most recent portion of the aperture where phases of the I & Q samples are adjusted to mimic a correction in motion measurements. The data are then processed as usual with backprojection to ensure well-focused video frames, even during periods of severe turbulence, changes in GPS satellite sequencing or IMU errors.