Non-uniform texture migration method and device

Abstract

According to a non-uniform texture migration method and a device, the synthesis quality of texture migration is improved, and particularly, the color richness and distribution characteristics of textures are kept consistent with sample textures; more complex texture images are processed, so that the texture migration adaptability is wider, a user can more intuitively control the synthesis process, and the result of freely controlling the texture migration is realized. The method comprises the steps of through an input sample and a group of semantic annotation graphs, generating a target texture graph through the following steps: (1) texture structure extraction: extracting texture features through a user supervision mode, and extracting color and distribution features from an input texture sample; and (2) multi-scale iterative texture synthesis: integrating the color and distribution characteristics of the texture into an optimization-based texture synthesis method, constraining a texture migration target function by utilizing the characteristics based on the optimization-based texture synthesis method, and synthesizing the texture by utilizing a voting step through the target function by iteratively searching similar image blocks, and finally, generating high-quality non-uniform textures meeting user requirements.

Description

technical field [0001] The invention relates to the technical fields of computer graphics and computer image processing, in particular to a non-uniform texture transfer method and a non-uniform texture transfer device. Background technique [0002] In recent years, texture transfer has been widely used in various image editing scenarios, including image reconstruction, image inpainting, image compression, video synthesis, and image and video stylization and other application scenarios. Although texture images can be obtained by many methods, such as manual drawing or photography. But when it comes to production, you need to use the same large number of textures. At present, sample-based texture transfer is still one of the most popular methods, because this method can handle multiple types of textures, and is easy to use. Users only need to provide a sample to generate high-quality texture results. The composite result can be of any size and will not contain too many repet...

AI Technical Summary

Problems solved by technology

At the same time, only a single sample cannot achieve high-quality composite effects

Images