The fault-karst reservoir, which evolved from the deformation and karstification of carbonate rock, is one of the most important reservoir types in Western China. Along the deep-seated fault zones, there are a lot of widely spread and densely distributed fractures and vugs. The energy of the diffractions generated by heterogeneous structures, such as faults, fractures, and vugs, is much weaker than that of the reflections produced by continuous formation interface. When using the conventional full-wavefield imaging method, the imaging results of continuous layers usually cover small-scale heterogeneities. Given that, we have used a plane-wave destruction filter to separate the diffractions from the full data and image the separated diffractions using the least-squares reverse time migration (LSRTM) method. We use several numerical examples to demonstrate that the newly developed diffractions using LSRTM can improve the definition of the heterogeneous structures, characterize the configuration and internal structure of the fault-karst structure well, and enhance the interpretation accuracy for the fault-karst reservoir.