Image fusion based on convolution sparse representation and pulse coupled neural network in non-subsampled contourlet domain

In sparse representation, image data can be described as a linear combination of basis function. The sparse representation of image data is sparsely described in data blocks, disturbing the continuity between the data blocks, causing coding redundancy and blurring of details. Using a convolutional sparse representation, the image can be sparsely coded in its entirety, and the image sparse coding is performed by replacing the product of the coding coefficient and the dictionary matrix by the convolution sum of the characteristic response and the filter dictionary to achieve an optimised representation of the entire image. In view of the above defects, this paper studies a fusion technique based on convolutional sparse representation and NSCT-PCNN (abbreviated as NSCT-CSR-PCNN fusion algorithm) and uses it in the image processing field. The algorithm uses the alternating direction method of multipliers (ADMM) instead of the orthogonal matching pursuit (OMP) algorithm to perform sparse approximation of low frequency sub-band to obtain the characteristic response coefficients and complete the fusion of low frequency sub-band. Experimental results show that the fusion effect of NSCT-CSR-PCNN algorithm is better than other algorithms. The fusion image has good visual effect with clear texture, high discrimination and high contrast.