The Proceedings of Mechanical Engineering Congress, Japan
Online ISSN : 2424-2667
ISSN-L : 2424-2667
2015
Session ID : J2410203
Conference information
J2410203 Reconstructing Space-Variant Blurred Images of Endoscope with Superpixel Segmentation
Kenshi OTSUZUMIYosutoshi ISHIHARA
Author information
CONFERENCE PROCEEDINGS FREE ACCESS

Details
Abstract
The space-variant blurring (zoom blurring) occurs due to the endoscope's movement to the depth direction. The intensity and direction of the zoom blurring varies with the position of the image. Therefore, it is unable to apply the restoration method for space-invariant blurred images. In this paper, in order to approximate the zoom blurring as space-invariant blurring, the image including zoom blurring is divided into small regions (superpixel) where the blurring property become equivalent. Therefore, the restoration image can be obtained by removing the blurring in each superpixel with a general restoration method for space-invariant blurring. In order to confirm the validity of this proposed method, we carried out the computer simulations and the real image experiments.
Content from these authors
© 2015 The Japan Society of Mechanical Engineers
Previous article Next article
feedback
Top