Minimal Similarity Accumulation Attribute Using Dimensionality Reduction with Feature Extraction

In this abstract, we present an improvement in the Minimum Similarity Accumulation (MSA) attribute. The MSA is an efficient approach to highlight discontinuity regions in a seismic volume. We have extended the previous method by using Principal Component Analysis and Restricted Boltzmann Machine in order to reduce the dimension of the input volume preserving relevant features. This approach attenuates the noise interference and significantly improves the quality of the final attribute. As shown in the results, relevant features such as fault and channels are well outlined in the seismic image. The proposed attribute was evaluated on seismic blocks from an offshore region of New Zealand. Non pre-conditioning was applied.