In the three-dimensional (3D) shape measurement based on structural light fringe projection, various factors due to such as environmental environment can produce point cloud noise. In the existing point cloud denoising method, the point cloud denoising needs to be carried out by analyzing the geometric relationship between 3D space and point cloud, and is faced with a series of problems such as complex calculation and low efficiency. In order to improve point cloud accuracy and denosing speed, a point cloud denoising method is proposed based on image segmentation. Firstly, a 2D point cloud map image is established according to the 3D point cloud reconstructed by structured light fringe projection. Secondly, segmentation of 2D point cloud map image is performed by using threshold segmentation method and region growing method. Then, the segmented noise area is recorded and removed. Finally, the new 2D point cloud mapping image is re-reconstructed in 3D space to obtain a point cloud with noise removed. Experimental results show that point cloud accuracy can reach 99.974% after the proposed method, and denoising time is 0.954 s,which can effectively remove point cloud noise and avoid complex calculations in 3D space.