Ultrasonic　testing　of　coarse-grained　materials　is　strongly　influenced　by　high-level　scattering　noise.　In　addition,　the　signal-to-noise　ratio　(SNR)　and　spatial　resolution　of　imaging　by　the　traditional　total　focusing　method　(TFM)　are　relatively　low.　In　this　study,　we　focused　on　the　reconstruction　of　high-resolution　ultrasonic　images　from　full　matrix　capture　datasets.　A　weighted　TFM　image　by　combining　the　inverse　problem-based　method　and　traditional　TFM　is　proposed　to　detect　defects　in　coarse-grained　steel.　The　proposed　method　was　used　to　image　defects　with　the　full　matrix　data　obtained　through　simulations　and　experiments.　The　simulation　and　experimental　results　show　that　the　weighted　total　focusing　method　can　significantly　improve　the　SNR　of　ultrasonic　imaging　in　coarse-grained　steel　and,　moreover,　it　can　improve　the　resolution　of　imaging　and　distinguish　adjacent　defects　with　a　centre　distance　less　than　the　Rayleigh　criteria.