Summary To obtain subsurface images/reflectivity, a sequential workflow of full waveform inversion (FWI) and least-squares reverse time migration (LSRTM) is often used. With the vector reflectivity-based acoustic wave equation, we can simultaneously obtain a high-resolution velocity model and the corresponding impedance-derived reflectivity image. However, the inversion process is quite non-linear and the field data is often blurred by noise, which makes the inversion process challenging. In this work, we propose an implicit FWI imaging (IFWIM) workflow, where the velocity and impedance models are implicitly represented by the weights of a neural network and can be resampled from the neural network at any desired resolution or even on irregular grids. The reflectivity components are then computed from the inverted impedance. The synthetic and field data examples show that the proposed method can recover high-resolution velocity models and reflectivity images as an effective way to perform joint imaging and velocity inversion.
Wang et al. (Sat,) studied this question.