本项目结合集成电路封装的检测需求，研究微聚焦X射线图像TV正则化混合去噪方法。包括：.(1) 微聚焦X射线图像混合去噪目标函数的建立：分析微聚焦X射线成像机理，建立符合成像机理的含混合噪声图像模型；在MAP框架下，基于TV正则化建立混合去噪的目标函数。.(2) 基于TV正则化的微聚焦X射线图像混合噪声去除算法：针对不同的混合去噪目标函数，研究混合噪声去除数值算法设计。重点解决复杂数据保真项的处理，正则化参数的自适应调整以及算法的求解速度问题。.(3) 基于字典学习和TV正则化的微聚焦X射线图像混合噪声去除：在TV正则化求解的基础上，结合集成电路图像的稀疏性先验知识，建立基于字典学习的混合去噪目标函数，研究适于集成电路微聚焦X射线图像的混合去噪算法。.本项目提出的混合噪声图像模型、基于TV正则化和字典学习的混合去噪方法，对高精度CT的精确恢复和快速重建具有重要的理论意义。

This project studies Total Variation(TV) regularization algorithms for micro focus X-ray images corrupted with mixed noise in Integrated Circuit(IC) packaging, including:.(1) The mechanism of micro focus X-ray imaging is analyzed and the model of microfocus X-ray images corrupted with mixed noise is established. Based on the mixed noise model and Maximum a Posteriori, the object functions for denoising the images corrupted with mixed noise are built with TV regularization..(2) TV numerical algorithms for denoising mixed noise are studied. This research focused on solving complex fidelity term, adjusting the regularization parameter adaptively and the algorithms’ speed..(3) Based on the sparsity constraints of IC images, dictionary learning term is introduced into TV regularization object functions. The adpative denoising algorithms based on dictionary learning are studied for solving new object functions..The studies on denoising micro focus X-ray image based on TV regularization and dictionary learning are important to high resolution computed tomography.