一种基于纹理方向场的图像分割方法

An Image Segmentation Algorithm Based on Oriented Texture Field

提出了一种基于纹理方向场的图像分割方法。根据图像平滑滤波原理构造了高斯滤波器与分块滤波器相结合的平滑方法,利用纹理梯度方向场特征形成纹理梯度图像,以此为基础用分水岭算法对图像进行分割。结果表明,将该方法应用于粘连大米颗粒图像,能有效分割粘连大米图像,较好地解决了分水岭算法的过分割问题,具有较强的边缘稳定性。

In this paper, a segmentation method is proposed by making use of Oriented Texture Field. According to the theory of smoothing filter, a smoothing way that gauss is combined with blocking filter is obtained, and the feature of oriented texture gradient is produced based texture gradient image, then the segmentation of image is made by using watershed algorithm. Results of experiment show that the introduced method can segment overlapping rice grains images; over-segmentation can be overcome greatly, and has more stabilization to edge.