対数振幅スペクトルを用いたPSFの推定とボケ画像の復元

抄録

In this research, it is assumed that the degraded images, which are represented as convolution of the original images with the PSFs (Point Spread Functions), are defocused images by the lens system or the limit of resolution. It has been proposed to estimate both of the PSF and the original image using a Bayesian-based iterative method. In the Bayesian-based iterative method of image restoration, we need to set suitably an initial value of the PSF and the original image. The initial value of the original image is a blurred image, since there is only a little difference between blurred and original images. On the other hand, as there is insufficient information on the PSFs, it is too difficult to set the initial values of the PSFs. If blurred images are represented as convolution of the original images with the PSFs, the band-limit of the blurred image will approach to a band-limit of the PSF. Therefore, we determine the initial value of the PSF by estimating the band-limit of the PSF from logarithmic amplitude spectrum of the blurred image. Finally, we confirmed the effectiveness of the proposed method by simulations and experiments with out-of-focus pictures.