Separation of reflection and transparency using epipolar plane image analysis

Authors:
Thanda Oo;Hiroshi Kawasaki;Yutaka Ohsawa;Katsushi Ikeuchi
Affiliations:
Department of Information and Computer Science, Saitama University, Saitama, Japan;Department of Information and Computer Science, Saitama University, Saitama, Japan;Department of Information and Computer Science, Saitama University, Saitama, Japan;Institute of Industrial Science, The University of Tokyo, Tokyo, Japan
Venue:
ACCV'06 Proceedings of the 7th Asian conference on Computer Vision - Volume Part I
Year:
2006

Citing 3
Cited 1

Separation of Transparent Layers using Focus

International Journal of Computer Vision
Using known motion fields for image separation in transparency

Pattern Recognition Letters
Polarization-Based Decorrelation of Transparent Layers: The Inclination Angle of an Invisible Surface

ICCV '99 Proceedings of the International Conference on Computer Vision-Volume 2 - Volume 2

Accurately estimating reflectance parameters for color and gloss reproduction

Computer Vision and Image Understanding

Quantified Score

Hi-index	0.00

Visualization

Abstract

The effect of reflection and transparency is superimposed in many real world scenes, which is caused by glass-like shiny and transparent materials. The presence of such incidental effect in a captured image has made it difficult to apply computer vision algorithms and has lead to erroneous results. Moreover, it disturbs the texture acquisition of the real-world scene. This paper presents an optimal method for the separation of reflection and transparency components. The method is based on the Epipolar Plane Image (EPI) analysis. The method is not like the ordinary edge-based EPI analysis, but instead it is an edge and color-based EPI analysis. To demonstrate the effectiveness of our method, we present the results of experiments using synthesized and real images which include indoor and outdoor scenes, from which we successfully extracted the reflection and transparency components from the input image sequences.