Problems of image restoration in metallography
In optical inspection and in metallography in particular it is necessary to reduce to minimum distortions on an image so to obtain highest possible measurement preciseness measuring object size on its image. One among methods to solve this problem is image restoration. Difficulty of image restoration is stipulated for following factors: PSF of the device is not determined a priori; PSF may have complex structure (non-linear and spatially non-invariant). Regarding to this, the division of image restoration task into several stages, is proposed: 1. Image fragmentation into areas where PSF is linear and spatially invariant 2. PSF estimation in all image areas. 3. Formulation and solution of reverse task of the original image restoration in the proposal that PSF is determined approximately. 4. Image reconstruction (composition). Difficulties of each stage: 1. Presence of essential geometric distortions. Since the model of distorting system supposed to be linear and pulse characteristic is spatially invariant, so real blurred and, probably, noisy image should have blureness degree uniform within all fragment. By this, the fragment is to include objects with contrast differences of brightness. 2. PSF estimation could be obtained by a priori method (using test body) or aposteriory method (using blurred image). By aposteriory method of PSF determination the main difficulty is the correct choice of PSF restoration window size: at one hand, not more than one border should get into the window, at the other hand, the noise should not cause false border detection.3. Main difficulty in realization of fragment restoration algorithm is the choice of the inverse problem regularization (IPR) parameter. The procedure of automated selection of IPR parameter by minimum artifacts criteria, is realized. 4. The difficulty of the composition stage is the necessity of restoration of fragment size upto initial, as well as exclusion of visible borders between fragments. This task is solved by addition of compensation frames to each fragment. All algorithms described are realized in MatLab environment.
optical inspection metallography Sufficient geometric distortions fragment restoration algorithm
Vladimir V. BYKOV Аlexey А. MASLOV Mikhail V. FILINOV Andrey S. FURSOV
JSC SPECTRUM-RII, Moscow, Russia
国际会议
第十七届世界无损检测会议(17th World Conference on Nondestructive Testing)
上海
英文
2253-2260
2008-10-25（万方平台首次上网日期，不代表论文的发表时间）