Bidirectional Image Resampling Method for Blurred Region Detection

Abstract

This paper presents a novel method for blurred region detection. The input image is divided into blurred and unblurred regions. The bidirectional image resampling technique is used to distinguish between the blurred and the unblurred regions. In experiment, we test on 704 low DOF images and compare to three previous methods. The results show that the proposed method provides the best segmentation results. The average value of accuracy is 0.8339.