A fusion algorithm based on composite decomposition for PET and MRI medical images

Medical image fusion is of great significance in medical analysis and disease diagnosis. Until now, there are still limitations to the current fusion methods to our current knowledge, such as the restricted ability to extract detailed information and the luminance deterioration. In this paper, we propose a composite decomposition algorithm to overcome the limitations mentioned above, and which consists of the following steps. Firstly, an improved structure tensor (IST) is proposed to convert the MRI images to the smoothing layer, edges layer and corners layer, which ensures the resulting image has better brightness quality; Secondly, Non-Subsampled Shearlet Transform (NSST) is adopted to decompose the smoothing layer of MRI and the Y component of PET obtained from YUV transform into low- and high-frequency sub-bands images; The use of NSST can allow important detailed information to be preserved in the resulting image. Thirdly, the weighted averaged fusion rule guided by gradient and local energy (WAGLE) is introduced to fuse the low-frequency sub-bands, which can further improve the brightness of the fused images. The Whole Brain Atlas Database is used to evaluate the proposed method. The experimental results show that the algorithm proposed in this paper has significant improvement in the smoothness, color brightness, edge and local features of the images. Qualitative and quantitative analysis results show that the presented method has overcome the disadvantages of some current methods, which is superior to the other seven advanced fusion methods.