摘要： 在许多采用马尔可夫随机场模型的基于样图的纹理合成算法中,邻域大小决定着这些合成算法的纹理合成质量和合成速度。但目前该参数是由用户根据目测的纹元大小进行试探性的设置,往往需要多次试探才能获得较为理想的邻域大小,设置起来非常不便。文章运用统计分析的方法,提出了一种邻域大小自动识别的算法。实验表明,在一个较宽的纹理范围内,该算法能够自动地识别出理想的邻域大小。

关键词: 纹理合成, 邻域, 阈值, 自动识别

Abstract: For many sample-based texture synthesis algorithms using markov random field model, the size of neighborhood determines the quality and speed of texture synthesis. However, the parameter is set up tentatively at present, based on the size of texels perceived by user. It is a trouble to gain the perfect size of neighborhood as many times tentative needed. This paper presents an algorithm for automaticallly recognizing the size of neighborhood by statistic and analytic approach. The results of test show that the algorithm can automatic recognizing the size of neighborhood perfectly for a wide range of textures.

中图分类号: 

• TP391.4