Efficient block adaptive point spread function estimation and out-of-focus image restoration scheme

In this paper, a blurred image restoration scheme is proposed. The blurring process is modeled as the convolution between a sharp image and a point spread function (PSF). Instead of deriving one PSF applied indiscriminately to restoring the entire image, the proposed scheme is block adaptive and distinct PSF is estimated in each block containing high frequency components, e.g. edges. Since only blurred edges are refurbished in the proposed scheme, the computing complexity can be largely reduced. After PSF estimation, the deconvolution process is performed by using Wiener Filter. Finally, a median filter is employed to mitigate the ringing artifacts. Experimental results show that the proposed scheme can enhance the PSNR value of the restored image by 2~3dB. The computing time, however, is much shorter than prior arts with similar performance.