摘要
基于TV/L2模型的总变差最小化方法在图像分解过程中受到正则化参数的影响。本文通过分析发现,当不同于经典模型而设置较小的正则化参数时,分解的图像特性发生了重要改变。据此,本文提出了一种新的基于TV/L2模型的双极性图像细节分解方法,该方法在正则化参数较小的情况下将观测图像分解为一个近似图像分量和两个具有正、负不同极性并反映不同信息的细节图像分量。将这种新的图像分解方法应用于印刷电路板CT图像的处理中。实验结果表明,通过综合利用本图像分解方法得到的图像细节信息,可以在有效抑制金属伪影的同时增强PCB图像中的有用信息。
The Total Variation （TV） minimization method based on TV/L2 model is affected by the regularized parameter in image decomposition, This paper suggests that when the value of the chosen regularized parameter is small and different from classic methods, the property of the decomposed image will have an important change. According to this, a new method for image bipolar detail decomposition based on the TV/L2 model is proposed in this paper. The new method can decompose an image into an approximate image and two detailed images with different properties, when the value of the regularized parameter is small. Applying the decomposition model to Printed Circuit Board （PCB） CT images, we found that by combining the information in the detailed images, an enhanced image with metallic artifacts being effectively reduced can be acquired.
出处
《CT理论与应用研究（中英文）》
2013年第1期75-83,共9页
Computerized Tomography Theory and Applications
基金
国家高技术研究发展计划"863计划"(2012AA011603)
国家自然科学基金项目(10971228)
