Structural　defects　in　self-lubricating　sliding　bearings　would　lead　to　local　stress　concentration　and　service　life　reduction.　The　absence　of　accurate　detection　technology　for　micro　defects　in　self-lubricating　coating　has　seriously　limited　their　application　in　mechanical　equipment.　A　total　variation　(TV)　fusion　terahertz　imaging　method　is　proposed　toward　the　unidentifiable　micro　defect　induced　by　overlapping　terahertz　echoes.　First,　a　range　of　indicators　is　developed　to　quantify　the　alterations　of　signal　characteristics　in　different　stages　of　the　defect.　Subsequently,　the　TV　fusion　based　on　these　indicator　images　could　clearly　identify　different　defects.　After　that,　a　multinomial　regression　model　is　established　through　the　mathematical　relation　of　defect　thicknesses　and　indicators,　thus　the　defect　thickness　could　be　obtained　accurately.　The　experimental　results　for　surface　defect　detection　and　internal　delamination　measurement　demonstrate　the　high　accuracy　and　excellent　robustness　of　the　proposed　method,　making　it　attractive　for　defect　detection　and　quality　assessment　of　self-lubricating　coating.