Grain transferring method based on multiple drawings

Abstract

The present invention is a texture transfer method based on diverse graphs. This method first divides the brightness value given by the target image into several levels with the participation of the user; then, uses a block-based method to synthesize textures using different texture samples for different brightness levels; finally, for the synthesized block, calculate the brightness difference of each texture block, if it is greater than the threshold value input by the user, subdivide the block into small blocks until the brightness difference in the subdivided blocks is smaller than the threshold value input by the user. For these subdivided blocks, an L-shaped neighborhood search method is used for texture synthesis. The invention is characterized in that the texture is synthesized from coarse to fine, the texture transfer speed is fast, the user can control the whole process, and the purpose of generating a new texture image according to the known texture image is realized.

Description

technical field [0001] The invention belongs to the technical field of computer virtual reality, and specifically relates to a method for generating a new texture image by using an existing texture image, which is used for constructing and drawing a realistic virtual environment. Background technique [0002] In the real world, there are very rich texture details on the surface of objects, and people can distinguish various objects with the same shape according to them. Therefore, the simulation of object surface texture details plays a very important role in realistic graphics rendering technology. How to effectively express these texture details in computer-generated graphics has always been a hot issue in the research of computer graphics and virtual environment construction and rendering technology. [0003] There are two main approaches to simulating texture detail on computer-generated objects. One is to model surface details with polygon...

AI Technical Summary

Problems solved by technology

[0004] In computer graphics, texture mapping technology has achieved great success, but there are three serious problems: first, the source of texture is relatively single

Hand-drawn pictures can meet artistic needs, but it is difficult to be as real as photos; and photos, as textures, are usually small and cannot cover the entire surface of the object

At this time, simple pasting will visually cause artificial traces

Second, there is no natural mapping from texture space to natural space, so the texture will be severely deformed when mapping

Third, texture mapping takes a lot of memory and bandwidth

Texture synthesis needs to solve two problems: one is modeling, how to estimate the texture generation process from a given limited texture sample image

[0045] Block-based texture synthesis is fast, but the synthesized image has limitations

The synthesis speed based on the L-shaped neighborhood search method is slow, but it is easy to control the synthesis process, and can synthesize new and unexpected texture images based on the sample image

The current texture transfer methods are all based on single-sample images, and the L-shaped texture synthesis method is used, and the synthesis speed is very slow.

Images