期刊名称：International Journal of Innovative Research in Science, Engineering and Technology
印刷版ISSN：2347-6710
电子版ISSN：2319-8753
出版年度：2016
卷号：5
期号：5
页码：7971
DOI：10.15680/IJIRSET.2016.0505231
出版社：S&S Publications
摘要：Microscopes are widely used to view magnified image of objects in different studies like morphologicalimaging, living cells viewing, geological survey and rock identification.While working with irregular object, imagesare often degraded due to uniform focusing and illumination problems. The main causes of image degradation are dueto different types of blur and noise introduced by different systems, devices or operations. The deconvolution is anoperation used in such image restoration to recover an image from an object that is degraded by blurring and noise.Blind deconvolution is the image restoration algorithm where the parameters responsible for blur are unknown. In ourproposed method, a series of digital images of the target object are acquired corresponding to different focal planes. Amultilayer neural network is trained with a known training dataset for which the 3D coordinates of target object areavailable. In this process, the neural network models blind deconvolution from training dataset .When presented withunknown sample, the trained neural network identifies focused pixels of all image stacks. The fully focused image isreconstructed from focus scanned image stack of target irregular object by combining focused pixels of all stackedimages. This also provides z-coordinates of pixels from the stack id for 3D viewing.
