摘要： 针对联合图像分解及边缘提取模型对复杂纹理图像提取的边缘信息不完整、存在奇异点等问题，提出一种改进模型。利用Weickert提出的扩散张量(International Journal of Computer Vision, 1999, No.2)对模型中边缘b的正则项进行改进，使其能分别在梯度方向和沿边缘方向控制边缘b的扩散率。数值实验结果表明，改进模型可提高边缘提取的准确性，降低奇异性。

关键词: 图像分解, 边缘提取, 梯度, 扩散率, 扩散张量, 偏微分方程

Abstract: Since the model combining image decomposition with edge extraction is not good enough in processing complex image, this paper proposes an improved model. The diffusion tensor by Weickert(International Journal of Computer Vision, 1999, No.2) is used to improve edge b regularization of the model, enabling respectively in the gradient direction and along the edge direction to control the edge b diffusion rate. Experimental results show that the improved model can improve the edge extraction accuracy, and reduce the singularity.

Key words: image decomposition, edge extraction, gradient, diffusion rate, diffusion tensor, partial differential

中图分类号: 

• TP911.73