Seismic imaging of the subsalt medium often contains invaluable information that can help locate the reservoir. However, a clear subsalt image highly relies on the accuracy of salt-body and subsalt velocity inversion, because both of them bear the kinematic information of seismic image at such a depth. Full waveform inversion (FWI) is a power technique that can enhance our ability of salt scenario interpretation, inspired by several successful applications. To achieve the goal of reliable salt-body/subsalt velocity inversion, we need to obtain not only a high-resolution interpretation of the salt structure, but also the low-to-medium wavenumbers of the velocity that is crucial to mitigate the notorious cycle-skipping issue. In this work we propose a reflection full waveform inversion (RFWI) workflow to improve the subsalt image. Instead of making FWI play the role of correcting for salt misinterpretation among some latest work, we are focused on implementing RFWI to the subsalt velocity building by making use of its ability of retrieving low-wavenumber model update along reflection wavepath. We introduce a RFWI workflow with a simultaneous optimization over the model perturbations and the low-wavenumber components, using a mixed scheme of objective function. A streamer dataset based on the benchmark model with massive salt body is used to validate our inversion strategy. We assume a reasonable interpretation of the salt geometry has been proceeded prior to our inversion, while there is little knowledge of the subsalt zone. In the situation of missing frequencies below 2 Hz and a limited offset-to-depth ratio below one, our RFWI workflow provides promising improvement in the subsalt imaging.