基于矢量扩展的纹理旋转修复

提出一种基于矢量扩展的纹理延伸方法,用于解决图像旋转后出现空白区域的问题.采用图像质量判断阈值来判断矢量化图像的相似度,将图像转换为矢量化的三次曲线组合;旋转矢量图以后,再根据矢量参数扩展空白区域曲线,并生长纹理堆积区域;最后用纹理渐变来填补矢量曲线和真实纹理的渐变差异.采用木纹图片做样本,实验显示本算法产生的目标图像高度接近真实木纹纹理.

Rotated Texture Image Repair Based on Vector Extensions

This paper presents an image repair method based on vector extension, which intends to correct the image blank area left by rotation. Using the image quality threshold to determine the similarity of vector image, the image is converted to vectors of cubic curve combination. After rotating vector-graph, the curves are expanded to the blank area by adjusting the vector parameters, and the texture accumulation areas are growed. The texture gradient is used to fill the differences of vector curves and real texture gradient. Using the wood texture images as samples, the experiments show the repaired images are highly similar to real wood texture.