Seismic data summarization with content-aware resizing

Abstract

Seam carving (SC) is a computer vision algorithm that resizes natural images by removing their least informative regions. We adapt SC as a novel application to reduce the size of 2D and 3D seismic data with amplitude preservation. Unlike decimation and conventional resizing, the proposed structure-aware reduction algorithm keeps the data's most important structures and textures. In practice, the SC method uses a gradient-based energy operator and dynamic optimization to find the optimal reduction of the seismic data. We introduce an energy function that uses Gaussian kernels of variable size to compute the magnitude of the data derivatives and implement a quantitative measure to compare different SC alternatives. The proposed Gaussian-based algorithm yields reduced seismic data sets that preserve the main structures and textures of the original data even in the presence of noise. The reduced data are not downsized versions of the original image or volume. We see the seismic summary as representative new data that can help interpreters and processors in gaining insight and assessing the results of filters and seismic attributes with fewer computational resources. Keeping this in mind, the proposed content-aware method is a valuable tool for assisting users in seismic data analysis and interpretation.