17:15 〜 18:30
[MGI33-P03] 岩石コア試料のX線CT画像の超解像とオマーンオフィオライトの蛇紋岩への適用
キーワード：Oman Drilling Project、Image Superresolution、 Sparse Representation
Recently, the X-ray computed tomography (CT) scanner has been commonly used for the onboard analyses of cores of scientific drilling, which provides us the 100 to 1000 m scale continuous data of various structures (lithological or sedimental layering, fractures, pores etc) of rocks. However, the resolution of such X-ray CT for the cores (for example, ~0.17 mm by the X-ray CT of Drilling Chikyu) is slightly low compare to reveal the shape and geometry of mineral grains and thin fractures. In contrast, the resolution of X-ray CT in usual laboratories is high (~0.01 mm or less). Therefore, if we can link the high resolution X-ray CT in the lab of the limited samples to the low resolution drilling core CT data, we can obtain the valuable big data for the analyses of multi-scale structures.
For this purpose, we have tried to apply sparse superresolution techniques to the rock samples with veins. The sample is a serpentinezed dunite in the Moho transition zone taken by the Oman Drilling project (CM1A). We took the X-ray CT images of the samples with the voxel size of 10, 20, 40 and 80 micrometers. The X-ray CT image reveals that distribution of magnetite and spinel grains and fractures filled with serpentine. We consider that each image is expressed by sparse representation; an image consists of a small number of basis images. The algorithm consists of two processes: dictionary learning and superresolution processes. In dictionary learning process, basis images are obtained by feature extraction from multiple X-ray CT images. In super-resolution process, a reconstructed high resolution image is obtained from a low resolution X-ray CT image by using common sparse representations. By using the sparse superresolution algorithm, we obtained results that sparse superresolutions provide fine structure from low resolution X-ray CT image.
For this purpose, we have tried to apply sparse superresolution techniques to the rock samples with veins. The sample is a serpentinezed dunite in the Moho transition zone taken by the Oman Drilling project (CM1A). We took the X-ray CT images of the samples with the voxel size of 10, 20, 40 and 80 micrometers. The X-ray CT image reveals that distribution of magnetite and spinel grains and fractures filled with serpentine. We consider that each image is expressed by sparse representation; an image consists of a small number of basis images. The algorithm consists of two processes: dictionary learning and superresolution processes. In dictionary learning process, basis images are obtained by feature extraction from multiple X-ray CT images. In super-resolution process, a reconstructed high resolution image is obtained from a low resolution X-ray CT image by using common sparse representations. By using the sparse superresolution algorithm, we obtained results that sparse superresolutions provide fine structure from low resolution X-ray CT image.