Block-based fast image restoration
Sungjun Yim, Jeong-Ho Shin, Joonki Paik · 2007
Fast Fourier transform (FFT) is powerful, fast computation framework for convolution in many image restoration applications. If the pixel values of opposite boundaries of the image do not match, the restored image is corrupted by interference, called boundary artifacts. We propose a new block-based fast image restoration algorithm using combined extrapolation and edge-tapering. We separate sub-blocks into inner blocks and out blocks. And we apply edge-tapering to the inner blocks and extrapolation to the outer blocks respectively. This procedure is shown to lead to best results than classical methods with reduced computational load.