As the resolution in coherent diffractive imaging improves, interexposure and intraexposure sample dynamics, such as motion, degrade the quality of the reconstructed image. Selecting data sets that include only exposures where tolerably little motion has occurred is an inefficient use of time and flux, especially when detector readout time is significant. We provide an experimental demonstration of an approach in which all images of a data set exhibiting sample motion are combined to improve the quality of a reconstruction. This approach is applicable to more general sample dynamics (including sample damage) that occur during measurement.
© 2011 Optical Society of America