Matrix Completion for Block-wise Missing Information Case in Subsurface Imaging

Missing information in subsurface imaging during field measurements is a common problem and frequently encountered block-wise missing information is the most challenging one. In this study, we apply the nuclear norm minimization (NNM) based matrix completion algorithm for the block-wise missing information case. This algorithm is proved its superiority compared to the other available methods for the column-wise missing data problem thus it is preferred as default matrix completion method. The critical relation between patch size and matrix completion performance in NNM is clearly shown for the block-wise missing information case. The detailed visual and quantitative analysis both on simulated and real data demonstrate the completion performance of the method for the extreme block-wise missing information cases.