The gear flank exhibits a smooth yet intricate geometry, and point cloud obtained through line-structured light frequently contain substantial outlier noise, and the point cloud reconstruction will have a certain position error. To tackle the challenges of accurately identifying and mitigating outlier noise, as well as correcting pose errors during point cloud reconstruction, this paper presents a denoising and registration methodology for gear measurement utilizing line-structured light. A local coordinate system is established based on the normal distance from the gear’s point cloud to its theoretical contour line, facilitating the determination of an effective denoising threshold for rapid and efficient noise removal. By leveraging involute properties, normal of the gear point cloud are computed to formulate a novel objective optimization function; registration is subsequently achieved via Gauss-Newton iteration. Both simulation and experimental measurements validate that the proposed method demonstrates superior performance compared to existing techniques.