Multimodal Medical Image Fusion Based on Multiple Latent Low-Rank Representation

Algorithm 1

Proposed image fusion algorithm.

 Input: and .

 Output: fused image

 / Part 1: layer multiple LatLRR decomposition. /

1 for eachdo

2 for eachdo

3  Perform LatLRR decomposition on to obtain and .

4 end

5 end

 / Part 2: Fusion of saliency part. /

6 for eachdo

7 perform max-rule on to obtain as euqation (8).

8 end

9 Superimpose to obtain the final fused saliency part as equation (9).

 / Part 3: Fusion of low-rank part./

10 for eachdo

11 Input to VGG-19 network to obtain extract from the 5th layer of the network;

12 Solve the -norm of to obtain activity level map as equation (10);

13 Calculate the initial weight map by equation (11);

14 Enlarge to get the final weight map as equation (12).

15 end

16 Calculate the final fused low-rank part as equation (13).

 /Part 4: Reconstruction./

17 Superimpose the fused saliency part and the fused low-rank part to obtain the fused image as equation (14).