Reflection tomography based on a velocity model with implicitly described structure information

Seismic reflection tomography is a poorly conditioned inverse problem. A standard tomography scheme is generally performed by adding a sparse parameterization of the model and/or regularization criteria to improve the morbidity. However, isotropic features are prone to creation in the obtained velocity model, which is often considered as geologically unrealistic. We improve the tomographic effect from two aspects: model parameterization and constraints incorporation. As for model parameters, a novel model representation method is proposed. In the model, velocity discontinuity is implicitly described as a zero-value contour of a signed distance field. All the velocity interfaces constitute a division of the entire model space and thus create a natural spatial framework for tomography. Under the constraint of this structural framework, common Tikhonov-type regularization items and other geological information are imposed easily and flexibly. Different model regions (blocks) are independent and can be inverted without mutual interference. The incorporated geological information functions locally and pertinently, and makes the inverted model be more geologically reasonable. The results from 2D field data are shown. The obtained velocity can improve the focusing of the migrated image.