融合变分和深度先验的非均匀图像分割算法

Heterogeneous Image Segmentation Algorithm Combining Variational and Depth Prior

为了尽可能精确地分割出强度不均匀图像更多的细节部分,提出一种融合深度图像先验的变分图像分割模型,利用交替方向乘子法设计相应的数值求解算法。实验结果表明,提出的模型在去噪正则化(regularization by denoising,RED)框架下融合了TV正则项捕获边缘和卷积神经网络(convolutional neural network,CNN)捕获细节的优势,尤其在处理结构丰富和纹理细致的图像时,可以分割出更多的细节,分割结果更精确。同时,提出的方法对于不同的初始轮廓具有很好的鲁棒性。此外,与对比实验中处理非均匀图像分割的方法相比,该模型算法复杂度低,具有快速高效的优势。

To segment more detailed parts of a homogeneous image as accurately as possible,a variational image segmentation model with depth image prior is proposed in this paper.The alternating direction method of multipliers(ADMM)is used to design the corresponding numerical algorithm.The results reveal that the proposed model by incorporating the denoising regularization(RED)framework with TV regularization can retain more details in images and obtain more accurate segmentation accuracy.At the same time,the model is robust to different initial contours.In addition,compared with the image segmentation method in the comparison experiment,the algorithm complexity of the proposed method is lower and has higher operational efficiency.