Multi-focus image fusion with Geometrical Sparse Representation

Highlights：

• This paper proposes a novel multi-focus image fusion algorithm based on Geometrical Sparse Representation (GSR) over single image, which shows the potential of geometrical sparse representation coefficients used for image fusion.

• This paper applies an average pooling strategy to determine the situation of the image patch and then obtains an initial decision map. The average pooling helps us take all the patches which involves the pixel k into consideration to make the decision more precise.

• This paper employs a weighted GSR model in the sparse coding phase, ensuring the importance of the center pixel. The average rank of six evaluation metrics demonstrate the proposed GSR outperforms state-of-the-art image fusion algorithms.