Partially blurred images restoration using adaptive multi-stage approach

This paper presents an efficient image restoration algorithm that is content adaptive. Traditional image restoration or de-blurring algorithms apply the same logic on the whole image even when only few image blocks are blurry. The developed algorithm is designed to use less calculations and save the un-blurred image parts quality during the de-blurring process. When more than one image is used, our algorithm is designed to select the best extracted quality from each image on a block bases. Our proposed approach consists of three main stages. In the first stage, the detection of blurry blocks of an image is performed. In the second stage, a partial de-blurring is applied only to the detected blurry blocks. In the third stage, an image fusion technique is used to fuse the de-blurred blocks back into the final image to be restored. MATLAB was used for our simulation results. Simulation results using various evaluation metrics showed that our proposed approach has achieved up to 14 dB (36%) enhancement based on PSNR for partially blurred images where blurry blocks are not in the same spatial position in the input images and close to 1 dB in cases where the blurry blocks are in the same spatial position.