Abstract: In traditional medical image fusion process based on single pixel, the information similarity is ignored and detailed information may loss. To solve these problems, an image fusion technology based on Nonsubsampled Contourlet (NSCT) and Robust Principal Component Analysis (RPCA) model with similar information is proposed. First, the initial matrix constructed by the image block from the original images is decomposed into low frequency and high frequency parts by NSCT transformation. Second, the low rank component is decomposed into low rank matrix and spare error matrix by using the low rank matrix model with similar information. Third, the low rank matrix of the two images, the spare error matrix and high-frequency components are fused by fusion rule of the absolute maximum method. Finally the fusion image is replaced by the inverse transform. The experiment results on CT and MRI show that the proposed method can maintain more edge and texture detailed information of the source images.

Key words: image processing, image fusion, Nonsubsampled contourlet transform(NSCT), robust principal component analysis, low-rank matrix