To　measure　the　surface　roughness　of　objects,　a　non-contact　measurement　method　based　on　shape　from　focus　was　proposed.　This　method　uses　sequences　of　captured　two-dimensional　images　to　reconstruct　the　three-dimensional　surface　shape,　and　calculates　surface　roughness　parameters　based　on　the　reconstructed　height　data.　The　accuracy　of　three-dimensional　reconstruction　depends　on　the　accuracy　of　the　initial　focus　volume　of　the　image　and　the　optimization　of　the　initial　depth　map.　To　solve　this　problem,　it　is　proposed　to　use　the　curvature-based　focus　measurement　method　to　calculate　the　focus　value　at　each　position,　and　classify　and　remove　the　noise　of　the　initial　depth　map　to　obtain　a　high-precision　reconstructed　depth　map,　thereby　ensuring　accurate　measurement　of　surface　roughness.　The　experimental　results　show　that　the　relative　errors　of　roughness　measurement　are　all　within　6%,　which　not　only　enables　evaluation　in　the　line　direction,　but　also　enables　comprehensive　evaluation　on　the　surface.　This　method　can　meet　the　requirements　of　measurement　accuracy　and　has　very　good　flexibility,　which　has　obvious　advantages　for　surface　shape　measurement　of　small-sized　targets.　　©　2024　IEEE.