2D/3D Image Morphing Technology from Traditional to Modern

Abstract

Image morphing technology is a crucial area in computer graphics and computer vision, enabling various effects such as transformation, rotation, and deformation of 2D and 3D images. It has become essential in the metaverse, virtual reality (VR), industrial design, and so on. In recent years, the emergence of artificial intelligence (AI) has brought significant changes to image-morphing technology, impacting deformation effects and technological development profoundly. A comprehensive introduction of image morphing strategy is provided from three perspectives: (1) Linear deformation methods based on control points of topological network structures, focusing on techniques such as free-form deformation, Laplacian deformation, mesh deformation, etc., and introducing and exploring the fundamental principles of these methods. (2) Nonlinear deformation methods based on conditional constraints, addressing the limitations of linear deformation by using constraints on position, volume, and area, as well as linear blend deformation method of dual quaternions, and discussing the technical limitations of nonlinear deformation. (3) Image generative morphing methods based on large neural network models, analyzing cutting-edge techniques of the Diffusion models, Generative Adversarial Networks (GANs), and Sora world model architecture. They can fuse information from different datasets or modalities, and learn hidden distributions to achieve image generation, style transfer, and so on.