Application of 3D image processing technology based on image segmentation in packaging design

Recent years, the methods to combine artificial intelligence technology with 3D image processing technology has become a hub for research in packaging design. Traditional 3D images are mostly produced by professional equipment, but this method is small in scope and high in cost, which does not meet the needs of most people. To solve the above problems, this study combines the mean shift algorithm with the confidence propagation algorithm, and obtains the confidence propagation-mean shift algorithm. In addition, the Lucascanard-confidence factor optical flow algorithm is improved by introducing the confidence factor to the Lucascanard-confidence factor optical flow algorithm. The research continues to combine the confidence propagation-mean shift algorithm with the Lucaskarnad-confidence factor optical flow algorithm to extract parallax maps and then synthesize 3D images. The results show that the iteration times and iteration time of the confidence propagation-mean shift algorithm are 9 times and 97.05 s, respectively. The number of parallax templates and the number of regions is 6 and 43 respectively. The confidence propagation-mean shift algorithm has 4 iterations, 36.8 s iteration time, 14 parallax templates and 65 regions in the category of portrait images. The accuracy of foreground depth, background depth and depth are 99.72, 99.87 and 99.80%, respectively, for the Lucas Kanard-confidence factor optical flow algorithm. In summary, the two algorithms proposed in this study have excellent performance, which can extract parallax map well and generate 3D image accurately, owning certain promotion value in the field of product packaging design.