Proceedings of the
9th International Conference of Asian Society for Precision Engineering and Nanotechnology (ASPEN2022)
15 – 18 November 2022, Singapore
doi:10.3850/978-981-18-6021-8_OR-11-0211
Parametric point-spread-function model optimization for microscopic image super-resolution
11 Department of Mechanical Engineering, National Taiwan University, No. 1, Section 4, Roosevelt Rd., Taipei City, 10617, Taiwan
2
ABSTRACT
Image resolution degrading problems can roughly be classified into two groups, diffraction limit and optical aberrations. In general, the image resolution is limited by the optical diffraction limit. Again, optical aberrations induced by the lenses can also lead to severe image blurring or quality degradation. The article presents a novel image deblurring algorithm by proposing a new parametric point spread function (PSF) estimation in a modified Wiener deconvolution process. The optical PSF is first modeled using a JINC-liked function by considering linear-spatial invariant aberrations and imaging parameters. An iterative Wiener deconvolution algorithm is used to optimize the quality of test images. The test image's coefficient of variance (CV) is set as the objective function for model optimization. Furthermore, to minimize undesirable ringing artifacts and noise amplification that may be generated in the above process, the gradient map density and CV of the image gradient map are proposed and developed for the regularization terms. Since pixel-wise deconvolution operation may suffer color distortion, a multi-channel parametric fitting method for image color correction is also applied. In addition to using JINC function as the PSF estimated basis model, we also take other types of function to model the PSF which introduce more freedom degrees for the modeling. By testing the developed method on some microscopic biomedical images, the experimental results show that the proposed method can effectively restore image detail information and improve the sharpness while maintaining the chrominance and chroma of the test images. As seen from the test as a preliminary verification, image super-resolution can be achieved by the developed method.
Keywords: Image superresolution, Point spread function, Image deblurring, precision metrology
11 Department of Mechanical Engineering, National Taiwan University, No. 1, Section 4, Roosevelt Rd., Taipei City, 10617, Taiwan
2
ABSTRACT
Image resolution degrading problems can roughly be classified into two groups, diffraction limit and optical aberrations. In general, the image resolution is limited by the optical diffraction limit. Again, optical aberrations induced by the lenses can also lead to severe image blurring or quality degradation. The article presents a novel image deblurring algorithm by proposing a new parametric point spread function (PSF) estimation in a modified Wiener deconvolution process. The optical PSF is first modeled using a JINC-liked function by considering linear-spatial invariant aberrations and imaging parameters. An iterative Wiener deconvolution algorithm is used to optimize the quality of test images. The test image's coefficient of variance (CV) is set as the objective function for model optimization. Furthermore, to minimize undesirable ringing artifacts and noise amplification that may be generated in the above process, the gradient map density and CV of the image gradient map are proposed and developed for the regularization terms. Since pixel-wise deconvolution operation may suffer color distortion, a multi-channel parametric fitting method for image color correction is also applied. In addition to using JINC function as the PSF estimated basis model, we also take other types of function to model the PSF which introduce more freedom degrees for the modeling. By testing the developed method on some microscopic biomedical images, the experimental results show that the proposed method can effectively restore image detail information and improve the sharpness while maintaining the chrominance and chroma of the test images. As seen from the test as a preliminary verification, image super-resolution can be achieved by the developed method.
Keywords: Image superresolution, Point spread function, Image deblurring, precision metrology
