Multiple Point Statistics with Pyramids Application on the Multi-scale Multi-structure Training Images
T. Chugunova, J. Straubhaar and P. Renard
Event name: Petroleum Geostatistics 2019
Session: Poster Session 1
Publication date: 02 September 2019
Info: Extended abstract, PDF ( 862.58Kb )
Price: € 20
Multiple Point Statistics (MPS) is now a well-known geostatistical method. In practice, one of the first operational need is to reproduce the heterogeneity with its small and large scales features, which are often present in natural phenomena. To respond to this need, multiple grid or flexible size template can be used. But unfortunately, even using these options, the large scale structure is not correctly reproduced or not reproduced at all. The human eye may make an abstraction from small scale texture and capture the large scale feature. Could the MPS approach be inspired by this idea to “see” large scale organization? A possible solution is to use a pyramidal representation of the Training Image similar to a Google Earth satellite image storage. This idea was implemented on the basis of the Multiple Point Direct Sampling algorithm (MPDS-pyramid) and this work presents its application to one synthetic and one real cases.