Deconvolution Inspection Technology for Optical Microscopy

• Main Authors: Wun-Mao Luo, 羅文懋
• Other Authors: Ming Chang
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/22428018761042526407

碩士 === 中原大學 === 機械工程研究所 === 99 === In the imaging formation of far-field optical microscopy system, for the limit of lens size and focal distance, there certainly is the loss of signal in the light diffuse from the object surface to the objective lens, this phenomenon is called as diffraction of optical system. Diffraction would blur the system images, when the sample size is nearly 1/2 of the wavelength of the light source, the sample image would be hard to identify. The limit is called the Abbe diffraction limit, at present, the Abbe diffraction limit is the thorny problem in the automated optical inspection system. The major purpose of this thesis is to combine the image processing and model-based deconvolution technology to recover the sample image, and expect the resolution of the recovered image of the optical inspection system that can break the diffraction limit. By the optical diffraction theory, we know when a single point light signal imaging through the microscopy system, it will spread as an Airy disc image, it also is called PSF (point spread function), and the imaging formation of optical microscopy can be seen as the convolution of the PSF. Therefore, we can restore the image by deconvolution with the true PSF, and to inspect the object small than the diffraction limit. In this thesis, we deconvolute the far-field optical microscopy images by Richardson–Lucy algorithm with the model-based PSF and the blind guess PSF, and use some image pre-processing technique to analysis and compare the result data. The system is confirmed that can inspect 100 nm linewidth object with the ideal conditions by computer simulation, we also prove the model-based deconvolution can inspect 200 nm linewidth sample with white-light source, NA0.55, and 100X Mirau interference objective, the resolution of the system is 74 nm/px, the inspection data for 200 nm is 3 px, the data for 350 nm is 5 px, the error for all data does not exceed 1px. It means the model-based deconvolution that can substantially improve the resolution of the degenerative inseption images, to attain the purpose for sub-pixel inseption precision and breaking the diffraction limit of the optical inspection system.