Image Reconstruction for Nondestructive Evaluation

• Main Authors: Huang Kuo Ming, 黃國銘
• Other Authors: Hung Hsien Sen
• Format: Others
• Language: zh-TW
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/19975503987535740073

碩士 === 國立海洋大學 === 電機工程學系 === 88 === Imaging flaws embedded in a structural object has become increasingly important in quantitative nondestructive evaluation (NDE). Tomographic images can provide very accurate information about the type, location, size, shape, and orientation of flaws. However, imaging an object by an X-ray tomographic scanning system can be difficult in NDE. The size of the object under test may be too large so that only projections within a limited-angle range can be obtained. Lack of complete angular coverge in CT scanning renders most of the Fourier-based image reconstruction methods, such as filtered back-projection (FBP), ineffective. As a result, they usually produce severe artifacts and also degrade accuracy in reconstructed cross sections. In this thesis, the CLEAN method and the method of projection onto the convex set (POCS) are investigated for image reconstruction in limited-angle problem. The method of CLEAN deconvolutes the known point spread function from an observed image and further improves image quality if a flawless prototype image is incorporated. However, the algorithm is rather time consuming due to slow convergence speed. To improve the convergence speed, a modified version of CLEAN is proposed. On the other hand, lack of Fourier values as prior information degrades the effectiveness of the POCS method. To overcome this problem, a modified POCS method is proposed. This method uses available projection data combined with the estimated projection data in the missing cone, incorporates a flawless prototype image into the algorithm design. As a result, the artifacts occurred in edges can be greatly reduced, the convergence speed is increased, and the quality of the reconstructed image is improved.