This　paper　proposes　a　method　to　diagnose　machine　faults　by　introducing　image　processing　technique.　In　this　paper,　the　machine　fault　diagnosis　is　considered　as　a　problem　of　Black　Casket.　Inputs　of　the　Black　Casket　are　voltage　excitation　signals,　and　outputs　of　the　Black　Casket　are　current　responses　signals.　In　order　to　deal　with　the　problem　of　Black　Casket,　a　health-condition　image　is　built　based　on　voltage　excitation　and　current　responses　of　the　machine.　Thus,　machine　faults　are　shown　in　forms　of　health-condition　image　deformations.　An　image　feature　detection　algorithm,　Maximally　Stable　Extremal　Region　(MSER),　is　used　to　detect　image　deformation　of　the　health-condition　image.　The　open-phase　fault　of　the　machine　is　studied　as　an　example　of　machine　faults.　The　effectiveness　of　the　proposed　method　is　verified　by　simulation　results.　©　2022　Division　of　Signal　Processing　and　Electronic　Systems,　Poznan　University　of　Technology　(DSPES　PUT).