Binary segmentation of volumetric images of porous media is a crucial steptowards gaining a deeper understanding of the factors governing biogeochemicalprocesses at minute scales. Contemporary work primarily revolves aroundprimitive techniques based on global or local adaptive thresholding that haveknown common drawbacks in image segmentation. Moreover, absence of a unifiedbenchmark prohibits quantitative evaluation, which further clouds the impact ofexisting methodologies. In this study, we tackle the issue on both fronts.Firstly, by drawing parallels with natural image segmentation, we propose anovel, and automatic segmentation technique, 3D Quantum Cuts (QCuts-3D)grounded on a state-of-the-art spectral clustering technique. Secondly, wecurate and present a publicly available dataset of 68 multiphase volumetricimages of porous media with diverse solid geometries, along with voxel-wiseground truth annotations for each constituting phase. We provide comparativeevaluations between QCuts-3D and the current state-of-the-art over this datasetacross a variety of evaluation metrics. The proposed systematic approachachieves a 26% increase in AUROC while achieving a substantial reduction of thecomputational complexity of the state-of-the-art competitors. Moreover,statistical analysis reveals that the proposed method exhibits significantrobustness against the compositional variations of porous media.
translated by 谷歌翻译