基于双参数颗粒分析的纹理分割

形态学颗粒分析是一种有效的颗粒纹理图像分割方法,但由于单一的颗粒度参数,使得其难以分割如颗粒度相同而空间排列不同的纹理图像,为此提出了一种基于双参数颗粒分析的纹理图像分割方法,该方法将传统的颗粒分析从单一参数扩展到以颗粒度和空间位置为参数的二维空间,使得扩展后的颗粒分析通过分布函数不仅能够提取纹理的颗粒度特征,而且能够获取纹理的空间排列特征,克服了传统颗粒分析难以区分颗粒度相同而空间排列不同纹理区域的问题.仿真实验表明,该方法在运算复杂度增加不大的情况下,纹理分割效果优于颗粒分析法,区域像素错分率低于颗粒分析法和Gabor滤波器法.

Texture segmentation based on double parameterized granulometry

Granulometry is an effective way for granule image segmentation. Since usual granulometry is based on a single size parameter and the spatial position feature of granule texture is not considered, it is difficuh to segment the texture image which contains different spatial arrangement objects with similar size and shape. Taking the spatial position factor into account, a method for texture segmentation based on double parameterized granulometry is proposed. Firstly, usual granulometry is extended to double parameter. Using the distribution function of double parameterized granulometry, more texture features are extracted. Secondly, fisher criterion is employed to degrade the feature dimension. Finally, classification criterion is given to segment texture regions. Simulations show that this method efficiently overcomes the shortcoming of usual granulometry for texture segmentation. Compared with usual granulometry and Gabor filter methods, the ratio of error classified pixels of proposed method is lower and the complexity is not evidently increased.