基于纹理特征的自然图像鉴别方法*

采用图像的纹理特性建立模型，实现了自然图像和人工图像（主要是简笔画、表格、函数图像）的鉴别。首先，计算两类图像的纹理特征，统计并分析数据找出规律；其次，选择图像的平滑度、一致性和图像熵作为代表性的纹理特征，组建数学模型；最后，利用三个特征的线性组合构成鉴别函数，并用模拟退火算法确定权值。实验表明，该方法可以很好地鉴别两类图像，识别效率较高。

Method for natural image discrimination based on texture feature

This paper aimed to establish a model based on texture features of image for discriminating the natural images and artificial images, such as mainly stick figure, form, function image. Firstly, computed two types of image texture features, and counted and analyzed the data. And furthermore, selected the representative texture features, such as smoothness, consistency and image entropy, and established the mathematical models. Finally, composed the differential function by a linear combination of the three features, and used the simulated annealing to determine the value of the weights. The experiment results show that the method can discriminate the two types of images effectively, and has higher degree of accuracy.