One of the most important branches of visual stylization is dot painting. It is primarily based on the change of point density to express the aesthetic features of the image by showing the change of color brightness in the image through the change of density. It is currently a hotbed of research in the topic of picture style migration. In many picture stylization applications, common depth learning methods have demonstrated considerable benefits, but they have not been applied in stippling. The fundamental problem is because stippling has a small dimension and constructing a loss function is challenging. A dot drawing generation algorithm based on superpixel and color knapsack is proposed in this paper. To begin, the algorithm uses a super-pixel preprocessing to preserve the color mutation level in the super-pixel, then generates the sampling radius using the color mean based on K-means binary clustering, and the Poisson disk generates the initial stipple sampling points based on the sampling radius. The random point selection algorithm based on the color knapsack algorithm is used to improve the local SSIM value to improve the effect of edge and detail area. The suggested technique outperforms previous methods in terms of visual effects, SSIM, and PSNR scores, as well as having good real-time performance.