Seismic image quality is highly dependent on the acquisition geometry, or more specifically, on the illumination of the subsurface. Ideally, the best image is obtained when each subsurface image point is illuminated equally, which potentially requires a recording geometry both finely sampled and infinite in length. In the real world, however, we always have a limited recording geometry. For example, in the standard marine-streamer acquisition system, only narrow-azimuth data with limited offsets are acquired. Recent developments in multi-azimuth or wide-azimuth acquistion techniques Howard and Moldoveanu (2006); Keggin et al. (2006); Michell et al. (2006) provide a richer coverage in azimuth, and the subsurface can be better illuminated, especially in subsalt areas. Much better images are obtained because of improved subsurface illumination. Nevertheless, the reality can never meet the requirement for infinite recording geometry and infinitly dense sampling for shots and receivers. Illumination holes and aliasing may still happen in complex geologies. Poor illumination results in poor images, and related artifacts destort the migrated image, making it difficult to interpret. This effect is readily visible in the reflection-angle and azimuth domain, where illumination holes and related artifacts can be identified. They are by no means random or weak, and thus simple stacking can not attenuate them effectively.
In this paper, I briefly review methods for extracting subsurface-offset-domain common-image gathers (SODCIGs) and angle-domain common-image gathers (ADCIGs). I demonstrate the effects of sparsely sampled wavefields on both SODCIGs and ADCIGs. I also demonstrate that the final image formed by simply stacking over the reflection-angle and azimuth axes without any weighting function suffers from artifacts caused by the poor illumination and has a low signal-to-noise ratio. Instead, I describe a simple but effective way to make the stacking process selective: we stack only those reflection angles and azimuths with good illumination. This method is tested on the wide-azimuth version of the SEG/EAGE salt data set, and a better image with higher signal-to-noise ratio is obtained.