The visual illusion is complementary to the art design techniques, which can be partially alienated according to this psychological characteristic of human beings and increase the interest in art design works. In this paper, after designing visual art images, the original local binary pattern is optimized to obtain an improved directional LBP algorithm to improve the accuracy and comprehensiveness of feature extraction. The K-means algorithm is used to cluster the colors extracted using the color space method into a color feature vector in the visual illusion art image. The empirical analysis is carried out from three aspects: feature vector similarity of visual illusion art images, image color clustering, and the application of visual illusion in the field of art design. The results show that: in the application example of clothing art design, the mean value of line spacing Z1=0.4cm is significantly larger than the line spacing Z2=0.8cm and smaller than the line spacing Z3=1.2cm in the line width Y1=0.4cm, which is larger than 1.25 and smaller than 0.09, respectively. Z3 is significantly larger than Z2, i.e., Z3>Z1>Z2. The mean value of Z2 (M=1.25) on the line width of 0.8cm is significantly greater than Z3 (M=0.8), indicating that lines with narrow width and wide spacing have the best slimming effect. The application of visual illusion in the field of art design can strengthen the effect of visual communication, enhance the interest of works, and improve the creativity and connotation of works.