Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:P1
Session:
Number:P1-20
Superpixel-Based Approach for View Synthesis Robust to Noise
Joo-Hyeon Kim, Tae-Youn Kim, Hong-Chang Shin, Jong-Il Park ,
pp.907-910
Publication Date:2016/7/10
Online ISSN:2188-5079
Summary:
Depth image based rendering (DIBR) technique allows to create a virtual view image by using images and disparity maps captured from different viewpoints. In general, the quality of the synthesized image highly depends on the accuracy of the disparity maps; therefore, the synthesized image has poor quality if noise exists in some region of the disparity map. For solving this problem, we propose a robust method to synthesize a high quality image of virtual view despite of disparity noise. The reference images are segmented into superpixels, and we use the disparity value which is not regarded as noise among disparity values in each superpixel for DIBR in order to suppress the effect of noise. The experimental results shows the effectiveness of the proposed method.