Abstract: The extraction method of image region of interest based on visual model possesses a distinctive superiority in extracting image region of interest area. Firstly, it fits human eye’s visual effect, so its outcome can express the visual’s real intention accurately in a large extent. Next, its procedure is more likely to achieve and its processing efficiency is much higher. The visual model based on Stentifold has become one of the presentations for its simple method and obvious effect. But it is also accompanied with the shortcomings that the results are  random, too sensitive to details and the time-consuming in calculating. For this reason, this papeer first attempts to adopt the method of combining histogram and uniform LBP, and make improvement of this method  which finally achieves an ideal effect.

Key words:  Stentifold model, histogram statistics, uniform LBP

摘要： 基于视觉模型的图像感兴趣区提取方法，在图像感兴趣区域提取领域，有着得天独厚的优势：首先，它贴合人眼视觉效果，因此可以较为准确地表示视觉的真实“意图”；其次，方法过程便于实现，处理效率比较高。基于Stentifold视觉模型的方法因其实现简单且效果显著而成为其中的众多代表之一，但其也存在结果随机性过大，对细节过于敏感，计算耗时的缺点。为此，尝试通过直方图统计、均匀局部二值模式（uniform LBP）结合的方法，对其进行改进，并取得了理想的效果。

关键词: Stentifold模型, 直方图统计, 均匀局部二值模式