During machining Al/SiCp composites, the machined surface takes on many defects such as grooves, pits, voids and protuberances, which greatly affect the characterization parameters of surface topography. However, a detailed investigation of the formation mechanism of surface defects in machining Al/SiCp composites is still lacking, and few surface roughness evaluation methods have been developed for separating the surface defects from roughness. In this paper, the formation mechanism of surface defects is revealed, and a roughness extraction method based on DT-CWT is established to extract the surface defects from roughness and conduct a roughness evaluation on the machined surface of Al/SiCp composites. Compared with the least-square method and Gaussian filter, the roughness extraction method based on DT-CWT has higher stability and can effectively distinguish between surface roughness and surface defects. According to the surface roughness evaluation based on the roughness extraction method, it is found that the surface roughness value of 2024Al/SiCp composite is lower compared to 6063Al/SiCp composite.