The detection of changes in sets of serial images acquired at different times is a common approach used in the context of underwater surveillance and monitoring. Coherent change detection with Synthetic Aperture Sonar (SAS) involves using both phase and amplitude of the data to detect very subtle scene changes that may not be detectable from the magnitude of the mean backscattered power. This paper briefly examines some of the parameters used during the co-registration process, which determine the shape of the warping surface used to resample the repeat-pass image onto the original image, using standard statistical metrics that can evaluate the quality of the image registration. Using the results of this analysis, a multi-look processing approach which produces a set of independent looks of the scene is proposed in order to enhance the underlying coherence of the scene and improve the detection performance over that of a single-look image. The technique is tested on a data set collected in the Mediterranean Sea over sand ripples using ATLAS Elektronik's Vision 1200 system, a 150 kHz SAS mounted on the SeaOtter Autonomous Underwater Vehicle. Results show that it is possible to detect sub-resolution coherent changes under these environmental conditions using the proposed approach, and a number of recommendations for processing repeat-pass SAS data are made.