Multimodal medical image fusion using residual network 50 in non subsampled contourlet transform

ABSTRACT Medical image fusion technology and its collective diagnosis are becoming crucial day by day. This task confers the latest algorithm for image fusion of medical images to many diagnostic complications. Firstly, transform is employed on input source images. The result of the application of transform is the decomposition of source images into various subbands. Eminent features are extracted from these subbands by using resnet50. These features are fused by phase congruency and guided filtering fusion rules. Finally, inverse transform gives the original image. The experiment results of this algorithm are compared with different methods by taking some pairs of medical images. Subjective and objective outcomes prove that the proposed algorithm exceeds the current methods by giving optimal performance measures in the area of medical diagnosis. Thus, it is revealed that the suggested multimodal image fusion model provides elevated performance over existing models via diverse diseases using MRI-SPECT and MRI-PET.