Two important aspects of surface engineering are measurement and analysis. While the acquisition of 3-D surface data is largely improved by optical profilometry, there has also been much research about their evaluation. Although the fractal-based or multiscale analysis of surface roughness is well known, these methods are not tailored for the characterization of the roughness of important machined surface textures produced by turning, grinding, honing, etc. In this work, a multiscale model of rough groove textures for 3-D surface measurements is introduced. Under the assumption of fractal groove profiles, the groove textures are modeled as superpositions of orthogonal ridge functions by the ridgelet packet transform. This provides a compact representation of groove textures. This allows a restoration of roughness components from accurate optical measurements, so that precise characterization can be achieved based on restoration. This method is validated for restoration of honing textures for real optical surface measurements on a cylinder liner.