The work aims to optimize the design of home soft decoration patterns with Chinese calligraphy and painting elements. On the basis of analyzing the characteristics and content of Chinese calligraphy and painting art elements, combined with the constituent factors and style characteristics of home soft decoration, composition, lines, coloring, and ink were applied to the optimization design of soft decoration. Subsequently, the Efficient Sub Pixel Convolutional Neural Network (ESPCN) based on convolutional neural networks was combined with the Laplace operator to form a Laplace ESPCN model for overall decorative pattern migration, and a local migration method based on Poisson fusion was proposed to improve it. The proposed overall decorative pattern transfer model still had a quality factor of around 0.5 when the noise density exceeded 50%, and the minimum peak signal-to-noise ratio could reach 40 dB. After combining it with the local transfer of Poisson fusion, the proportion of people in the target area segmentation was around 50%. The Laplace-ESPCN overall pattern transfer model proposed in the conclusion results in high image quality and good noise reduction effect. After combining with the local transfer method of Poisson fusion, the obtained image edges are smoother, which can be effectively applied in the optimization design of home soft decoration patterns and has certain promotion value.