Hybrid Coefficient Fusion in Multi-resolution Singular Value Decomposition Domain for Medical Images

Multimodal medical image fusion technology is an important branch of image fusion technology, which purpose is to integrate the effective features of multiple medical images to assist doctors to complete the diagnosis. In this paper, a new hybrid coefficient fusion in multi-resolution singular value decomposition (MSVD) domain for medical images is proposed. In the proposed model, the source images are decomposed into four sub-bands, the first low frequency sub-band adopts the fusion strategy of principal component analysis PCA algorithm, and the second low frequency sub-band adopts the guided filtering GFF image fusion algorithm to enhance the extraction of local features. Pulse coupled neural network is used to fuse the first and second high frequency sub-bands. Then, the fused luminance channel is obtained by applying the inverse MSVD transform to the fused sub-bands. The final fused result is computed by color channels of PET image and the fused luminance channel. Experimental results show that the proposed method is superior to other comparison methods in fusion performance and visual evaluation.