Poisson cloning using bilateral image coarsening

  • Authors:
  • Hao Wu;Dan Xu

  • Affiliations:
  • Yunnan University;Yunnan University

  • Venue:
  • ICIMCS '10 Proceedings of the Second International Conference on Internet Multimedia Computing and Service
  • Year:
  • 2010

Quantified Score

Hi-index 0.02

Visualization

Abstract

Poisson cloning can clone a source patch to target image seamlessly given a constant boundary condition, but it usually results in "bleeding artifact" near the boundary, which is difficult to prevent only by boundary optimization. To address the problem, we switch Poisson cloning to bilateral image coarsening (BIC) space that binds similar pixels together in smooth regions and leaves pixels across edges independent to reduce the bleeding artifact. We present three ways to project Poisson cloning into the BIC space. Direct projection projects a result image into BIC space by solving a minimization problem. Gradient projection projects boundary and gradient of a source patch instead of the cloned image, which is more robust and is more consist with the operation of Poisson cloning. Modified gradient projection computes a correction function with deleted gradient of boundary in BIC space, and it improves the computation performance while it can obtain the similar result as gradient projection when the regions near the boundary are smooth. Experimental results demonstrate that our method reduces the bleeding artifact significantly and the sacrifice of performance is quite small.