Learning Model for Restoring Blurred Images Based on High Frequency Components

Author : Tomio Goto, Tatsuya Mizuno

Abstract :In recent years, various methods have been proposed in the field of image restoration and many researches have been conducted actively. An important aspect of blurred image restoration is to restore high-frequency components such as object outlines. In this paper, we propose a model that focuses on the restoration of high-frequency components. The restored image using this method successfully restores high-frequency components and obtains higher PSNR than that of the conventional method for the entire dataset.

Keywords :Blind Image Restoration, Learning Method, MIMO-UNet.

Conference Name :International Conference on Control, Automation, Robotics and Vision Engineering (ICCARVE-25)

Conference Place Napier, New Zealand

Conference Date 20th Feb 2025