Title   name
2007년 KSIAM 봄 학술대회
 
  Title
  Speaker   Lee, Chang-Ock  
  Date 2007-05-26
  Host
  Place KAIST
  File  의 1 번째 Real Media 동영상입니다. 의 1 번째 강연자료입니다.
 
Abstract : A low level image processing, for examples, image denoising and finding a local structure inan image, is one of essential processes in computer vision. The nonlinear structure tensor [1]based on an anisotropic nonlinear diffusion processes has shown its capability for presentinguseful information of an image, such as homogeneous regions, edges, and corners. The diffusionprocess to obtain a nonlinear structure tensor is a natural extension Perona-Malik’s model in [2].We propose an unconventional nonlinear diffusion process to obtain a nonlinear structure tensorof an image. Since the nonlinear structure tensor initially has gradient information of an image,diffusion coefficients of PDEs based on image make sense to present a local structure in animage.We make such diffusion coefficients in a system of anisotropic nonlinear PDEs to obtainnonlinear structure tensor. We also prove the existence and uniqueness of proposed PDEs. Asuperiority of nonlinear structure tensor using diffusion coefficients based on image is shown inapplications, for examples, image denoising and corner detection.