Local Motion Blurred Image Restoration Based on the Reciprocal of DCT High-Frequency Mean Incremental Prior

Abstract

Aiming at the blur caused by fast-moving objects in surveillance video, we propose a new method to restore blurred images. First, we use the PiCANet saliency detection algorithm to obtain the saliency map of the blurred area. Then we use the saliency map to guide the soft segmentation algorithm to divide the blurred image into foreground and background layers. Second, we propose a reciprocal of DCT high-frequency mean incremental prior to constraining the solution space of clear latent images under the framework of maximum a posteriori probability. Finally, we also propose an improved bilateral filtering algorithm to enhance the details of the restored image. The experimental results show that our algorithm's deblurring visual effect and objective evaluation index are superior to other algorithms.