Medical fusion framework using discrete fractional wavelets and non‐subsampled directional filter banks

Image fusion in neuro diagnosis is intimidating due to its complexity. The heterogeneous natures of the original brain images make intermodal transmission difficult during fusion. Medical image fusion using complementary modalities results in loss of vital salient information. Poor fusion, colour deficiencies result due to similar processing for both the modalities. A dual technique is proposed using discrete fractional wavelet transform (FRWT) and non-subsampled directional filter banks for better extraction of salient image elements for improved diagnosis. The sparsity character of the coefficients FRWT is controlled by optimising the parity operator using Grey Wolf optimisation algorithm. Four sets of neurological multimodal magnetic resonance imaging and single photon emission computed tomography (CT) brain images are used from benchmark database for validation. The objective evaluation has been conducted using five metrics. The main values obtained from objective metrics based on the proposed technique are 6.3213 for Shannon entropy, mutual information is computed to be 2.7582, fusion factor is 1.9095, standard deviation is 0.1310, and edge strength is 0.76122 indicating improved diagnostic information and superior image quality. Subjective evaluation by a medico validates the findings with finer visual output and enhanced contrast in comparison with recent and state-of-the-art methods.