Abstract:

Aiming at the main problem in image fusion by single Principle Component Analysis (PCA) or Nonsubsampled Contourlet Transform, a novel image fusion algorithm based on 2DPCAPCNN transform is proposed. Firstly, 2DPCA is applied to multispectral image waveband. Its principle components are regarded as signal and its nonprinciple component is ignored because of being considered as noise. Secondly, the panchromatic mages and the first principle component are decomposed by nonsubsampled Contourlet transform (NSCT). The variance integration rules are used for the high frequent subband coefficients and the lowpass subband ones. Finally, the fusion image is achieved by inversing NSCT on the aforementioned subband coefficients. The simulation experiments show that the proposed method improves the spectral distortion as well as keeps the PCA transform good space resolution.

Key words: 2DPCA;NSCT;image fusion