一种基于自适应块拼接的纹理合成方法

纹理合成是当前计算机图形学、计算机视觉和图像处理领域的研究热点之一。本文在MRF模型和已有块拼接技术的基础上提出了一种自适应块拼接算法，该方法在全局最优的前提下进行不同大小纹理块的拼接，并使图像中的像素带有与视觉相关的权值，不仅在保证合成速度的同 时得到了更好的合成效果，而且从理论上统一了逐点合成和逐块合成两种基本方法。

A Texture Synthesis Method Based on Adaptive Patches

Texture synthesis is one of the hottest research topics in computer graphics, computer vision and image processing fields. We present a novel adaptive patch-based algorithm on the basis of the MRF model and other related algorithms. We make up patches with different sizes in a globally optimized way, and give the pixels of a image the vision-related weights. The new algorithm not only gets better results than other algorithms when the synthesis speed is guaranteed, but also unifies patch-based and pixel-based texture synthesis methods together.