Geometric-mean Reverse-time migration (GmRTM), a powerful cross-correlation-based imaging method, generates higher-resolution source images and is more robust to noise compared to conventional time-reversal imaging. The price to pay is the higher computational costs. Alternatively, we can adopt hybrid strategies by dividing the receivers into different groups. Conventional time reversal (i.e., wavefield summation) is performed inside each group, followed by the application of cross-correlation imaging condition among different groups. Such hybrid strategies can retain the advantages of both GmRTM and time-reversal, and are often more practical than pure GmRTM. Yet, designing appropriate grouping strategy is not trivial. Here, we propose two grouping strategies (adjacent and scattered) and use synthetic and field-data examples to evaluate their performance with various group numbers. In addition to the spatial resolution of the source image, robustness to random noise is another important assessment criterion, for which we consider two distribution patterns, such as concentrated and scattered, of traces contaminated with strong random noise. We also evaluated their effectiveness to visualize events (in the image domain) that are not completely recorded by all receivers. Our comprehensive tests illustrate the respective advantages of the two grouping strategies.
Diffraction imaging using geometric-mean reverse time migration and common-reflection surface