抑制纹理信息的偏置场变分图像分割模型

A bias field variational image segmentation model restraining texture information

偏置场变分水平集图像分割模型利用原始图像的局部灰度信息,可以对灰度不均匀图像进行有效的分割,但当灰度图像中存在纹理时,分割效果往往很差。针对这一问题,提出抑制纹理信息的偏置场变分水平集图像分割模型。利用一种基于纹理几何结构的纹理描述符描述图像中不同的纹理区域,使得不同纹理区域对比更加明显,相同纹理区域更加平滑,通过抑制纹理信息使后续的图像分割在纹理部分的错分大大减少。实验结果表明,相比偏置场变分模型,所提模型对自然及人工合成纹理图像均获得更好的分割结果。

The variational level set image segmentation based on the bias field can segment intensity inhomogeneity images using the local image information. However, the model cannot do it well when there are textures in the image. To solve the above problem, we propose a bias field variational level set image segmentation model to suppress texture information. The texture can be restrained by the intrinsic texture descriptor based on the texture geometric structure and it can enhance the contrast among dif- ferent texture regions and smooth the image in the same texture region. It can reduce the mistakes of texture region segmentation by restraining the texture. Experimental results indicate that the proposed model can better segment natural and synthetic texture images in comparison with the bias field variational model.