基于栅格元模型的纹理图像分割方法

提出了一种基于栅格元模型的纹理图像分割方法,该方法结合了视觉栅格元模型对不同纹理方向和频率的响应敏感性及特征变换的降维特性。首先通过栅格元运算模型提取图像的多维纹理特征,然后利用变换法将特征从多维空间映射到低维空间,最后依据降维后的纹理特征将图像分割为不同的纹理区域。仿真结果表明,该方法不但能有效地分割纹理图像,而且能分割出由纹理和非纹理区域组成的混合图像,与传统的Gabor滤波器法相比,分割效果更佳,适用范围更宽。

Texture Image Segmentation Based on the Grating Cell Model

A method of texture segmentation based on the visual grating cell model is proposed, which combines the response sensitivity of the grating cell model to different orientations and frequencies with the feature dimension reducing technique. First, texture features are extracted from the image via the grating cell computational model, and then the extracted feature vectors are mapped from the multidimensional space into one dimensional space through feature dimension reducing transformation, and finally, according to these one dimensional feature vectors, the texture image is segmented into several different texture regions. Simulations show that the proposed method can efficiently segment not only texture images but also mixture images which are composed of texture and non-texture regions, and segmented results are superior to the traditional Gabor filter method.