Optimizing the Input Model for Waveform Inversion Using Image-domain Wavefield Tomography with Illumination Compensation

Image domain wavefield tomography exploits focusing characteristics of extended images for updating the velocity field. In order to make good use of this information, we must understand how such images behave if the migration velocity is accurate. This is not trivial since focusing depends on the acquisition setup, as well as on illumination variation caused by the geology separating the acquisition array from the imaged structure, the data bandwidth, etc. We address this problem using a combination of migration/demigration to construct penalty functions that characterize focusing by incorporating acquisition parameters and data bandwidth. Moreover, instead of sampling the extended images at preset distance along the surface, we sample the image by constructing common image-point gathers, which are also much more economical from a computation point of view. Coupled with image residuals exploiting illumination-based penalty functions, we obtain robust wavefield tomography in areas of poor or uneven illumination. Models obtained with this type of methodology are good starting points to more sensitive, but less robust waveform inversion methods.