To better segment texture images in multiscales,we propose a hybrid model.This model uses the total variation model to control the multiscale representation of texture images,and introduces local self-similarity and local Gaussian distribution fitting functions as boundary extraction items,ultimately obtaining a hybrid multiscale texture image segmentation model.Experimental results show that the proposed model can segment texture images well in different scales,making the segmentation process more consistent with the perception mechanism of human vision.When the image is contaminated by Gaussian noise to varying degrees,the model can also accurately extract the target boundary.
关键词
纹理图像分割/多尺度/活动轮廓模型/变分水平集
Key words
texture image segmentation/multiscale/active contour model/variational level set
维普
万方数据