Multi-modality medical image fusion based on separable dictionary learning and Gabor filtering

Highlights：

• A multi-modality medical image fusion method based on a novel sparse representation (SR) so-called ASeDiL and Gabor energy in non-subsampled contourlet transform (NSCT) domain is presented.

• The ASeDiL-based method not only increases the sparsity decomposed by NSCT, but increases dimensions of texture extraction without adding dictionary redundancy, furthermore, the effective noise suppression performance of ASeDiL improves the accuracy of texture extraction.

• A popular special consistency based fusion approach is formulating an energy function, inspired by this, Gabor energy improves the spatial inconsistency of medical fusion image caused by sparse representation.

摘要

•A multi-modality medical image fusion method based on a novel sparse representation (SR) so-called ASeDiL and Gabor energy in non-subsampled contourlet transform (NSCT) domain is presented.•The ASeDiL-based method not only increases the sparsity decomposed by NSCT, but increases dimensions of texture extraction without adding dictionary redundancy, furthermore, the effective noise suppression performance of ASeDiL improves the accuracy of texture extraction.•A popular special consistency based fusion approach is formulating an energy function, inspired by this, Gabor energy improves the spatial inconsistency of medical fusion image caused by sparse representation.

论文关键词：Image fusion,Multi-modality medical image,Sparse representation,Gabor filter,Non-subsampled contourlet transform

论文评审过程：Received 19 September 2019, Revised 29 November 2019, Accepted 21 December 2019, Available online 7 January 2020, Version of Record 14 February 2020.