Detail-Preserving Multi-exposure Fusion for DR Images of Turbine Blades with Local Contrast Analysis and Exposure Intensity

Digital radiographic imaging technique based on the Digital Detector Array (DDA) plays an essential role in the non-destructive testing of turbine blades. However, the Digital radiography (DR) of turbine blade obtained by the X-ray spectrum at the single exposure parameter cannot provide effective information feedback to the whole blade in digital radiographic testing. Aiming at this critical issue, a detail-preserving fusion method of multi-exposure sequence DR images is presented in this paper. Firstly, the unordered DR images are divided into three categories of over-exposure, normal-exposure, and under-exposure by using clustering method. Then, the weight design methods based on local contrast analysis and exposure intensity are employed to accomplish the initial weight fusion. The guided filter and normalization are utilized to ensure the weight maps are edge-preserving and smoothing. Finally, the refined weight maps and initial DR images are decomposed by multiscale pyramid. The fused DR image can be reconstructed by the new fusion pyramid with satisfying visual effects. Four different experimental results show that the fused blade DR image can clearly show the full internal features of turbine blades, which integrally covers the advantages of high-exposure images and low-exposure images. Compared with the four existing DR fusion methods currently in use, the fusion effect of proposed method is most significantly improved. For four diverse kinds of turbine blades, all the fused images obtained by our method have the highest structural-similarity metric and the second shortest execution time.