基于局部边缘保持滤波器多尺度分解的工业射线图像增强  

Industrial ray image enhancement based on multi-scale decompositionof local edge preserving filter

作　　者：白强 桂志国[1] 刘祎[1] 张鹏程[1] 侯宏花[1] Bai Qiang;Gui Zhiguo;Liu Yi;Zhang Pengcheng;Hou Honghua(Shanxi Key Laboratory of Biomedical Imaging and Imaging Big Data,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学生物医学成像与影像大数据山西重点实验室,太原030051

出　　处：《国外电子测量技术》2022年第9期32-39,共8页Foreign Electronic Measurement Technology

基　　金：国家自然科学基金(61801438)项目资助。

摘　　要：在利用X射线检测复杂构造工件时,由于工件薄厚不均,会出现某些区域过曝光或欠曝光而导致细节丢失的情况。针对这一问题,提出了一种基于局部边缘保持滤波(LEP)的多尺度分解增强算法。首先,原始高动态图像经过对数化后通过LEP滤波器得到基本层,将对数化后的图像与基本层做差得到表示小梯度的细节层;然后将滤波半径逐步增大的LEP滤波器对基本层进行多次迭代滤波后做差得到多个细节层,对得到的多个细节层进行S型曲线映射后融合;最后将结果与原始高动态图像线性压缩后进行gamma变换然后归一化的结构层相加,得到的图像通过对比度受限的自适应直方图均衡化(CLAHE)的处理,便得到了背景与工件对比度很高且细节清晰的低动态范围图像。实验结果表明,提出的方法对复杂构建的工件具有明显的增强效果,并且算法效率很高。When X-ray is used to detect the workpiece with complex structure, due to the uneven thickness of the workpiece, some areas will be overexposed or underexposed, resulting in the loss of details. To solve this problem, a multi-scale decomposition enhancement algorithm based on local edge preserving filter(LEP) is proposed. First, the original high dynamic image is logarithmicized and the basic layer is obtained by LEP filter. The logarithmicized image is compared with the basic layer to obtain a detail layer representing a small gradient. Then, the LEP filter with gradually increasing filter radius is iteratively filtered on the basic layer for many times to obtain multiple detail layers. The obtained detail layers are fused after S-shaped curve mapping, The result is linearly compressed with the original high dynamic image and then added to the normalized structure layer by gamma transform. The obtained image is processed by contrast constrained adaptive histogram equalization(CLAHE), and a low dynamic range image with high contrast between the background and the workpiece and clear details is obtained. The experimental results show that the proposed method has obvious enhancement effect on complex constructed artifacts, and the algorithm efficiency is very high.

关 键 词：X射线图像 边缘保持滤波器 多尺度分解 CLAHE 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]