A real-time method for detecting sharp images in visual navigation

Abstract

In this paper, we present a novel method to filter the blurred images for improving the robustness of robot visual navigation to motion blur. Mainly arising from the motion blur of the robot, blurred images negatively affect the performance of visual navigation, and hence should be filtered. Based on the blur feature of the vision and the motion feature of the robot, our method, which we refer to as motion blur detector, is capable of outputting continuously sharp images for visual navigation system in a no-reference detection framework. Experiments are conducted to show the reliability and efficiency of the proposed method.