紀要論文

Split-and-Merge Algorithm for Motion Deblurring Based on DeblurGAN

eng

24321192

https://doi.org/10.15002/00021925

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

14

1

4

2019-03-31

Version of Record

Motion blur is a highly challenging problem in computer vision literature. Due to the ill-posed nature and the camera shake, the relative motion between the camera and the object in 3D space induces a spatially varying blurring effect over the entire image. This paper proposes split-and-merge algorithm based on DeblurGAN to make blurred image become relatively clear. For the same motion blur image, usually blurred degree of different parts is not the same. Therefore split-and-merge algorithm firstly split a blurred image into several parts and measure blurred degree for each part. Then the algorithm do deblurring for different times according to different blurred degree until all parts reach the acceptable degree. Finally the algorithm merge all parts into one image. The experiment results show this algorithm can help enhance sharpness of motion blur picture than original DeblurGAN.

Article

AA12746425