Time - lapse image registration using the local similarity attribute a

We present a method for registration of time-lapse seismic images based on the local similarity attribute. We define registration as an automatic point-by-point alignment of time-lapse images. Stretching and squeezing a monitor image and computing its local similarity to the base image allows us to detect an optimal registration even in the presence of significant velocity changes in the overburden. A by-product of this process is an estimate of the ratio of the interval seismic velocities in the reservoir interval. We illustrate the proposed method and demonstrate its effectiveness using both synthetic experiments and real data from the Duri time-lapse experiment in Indonesia. INTRODUCTION Time-lapse seismic monitoring is an important technology for enhancing hydrocarbon recovery (Lumley, 2001). At the heart of the method is comparison between repeated seismic images with an attempt to identify changes indicative of fluid movements in the reservoir. In general, time-lapse image differences contain two distinct effects: shifts of image positions in time caused by changes in seismic velocities and amplitude differences caused by changes in seismic reflectivity. The data processing challenge is to isolate changes in the reservoir itself from changes in the surrounding areas. Crossequalization is a popular technique for this task (Rickett and Lumley, 2001; Stucchi et al., 2005). A number of different cross-equalization techniques have been successfully applied in recent years to estimate and remove time shifts between time-lapse images (Bertrand et al., 2005; Aarre, 2006). An analogous task exists in medical imaging, where it is known as the image registration problem (Modersitzki, 2004). In this paper, we propose to use the local similarity attribute (Fomel, 2007a) for automatic quantitative estimation and extraction of variable time shifts between time-lapse seismic images. A similar technique has been applied previously to multicomponent image registration (Fomel et al., 2005). As a direct quantitative measure of image similarity, local attributes are perfectly suited for measuring nonstationary time-lapse correlations. The extracted time shifts also provide a direct estimate of Fomel & Jin 2 Time-lapse image registration the seismic velocity changes in the reservoir. We demonstrate an application of the proposed method with synthetic and real data examples. THEORY The correlation coefficient between two data sequences at and bt is defined as