Medical image fusion based on saliency and adaptive similarity judgment.

Information regarding single-modality medical images remains limited at present. Therefore, the details of relevant organizations cannot be reflected by such information and misdiagnosis may result in the clinical setting. A novel image fusion method based on nonsubsampled contourlet transform (NSCT) and visual saliency is presented for multimode medical images to solve the aforementioned problem and improve the effect of fused images. In this method, an image is initially decomposed via NSCT to obtain low- and high-frequency coefficients and derive the Achanta (AC) saliency map of the image. Subsequently, the similarity of low- and high-frequency subbands is calculated using the standard deviation and energy of a region. Similarity is determined through an adaptive threshold. Meanwhile, the saliency map is utilized to guide the fusion of high- and low-frequency coefficients, with emphasis on the visual salient target area of the source image. Lastly, inverse NSCT is used to reconstruct the fused image. The proposed algorithm can effectively improve the quality of multimodal medical image fusion and increase complementary information among different modalities as demonstrated by the experimental results. This algorithm is superior to existing medical image fusion algorithms and more in accord with human visual characters. [ABSTRACT FROM AUTHOR]