Haar wavelet transform-based optimal Bayesian method for medical image fusion
Jayant Bhardwaj,Abhijit Nayak
DOI: https://doi.org/10.1007/s11517-020-02209-6
Abstract:Image fusion (IF) attracts the researchers in the areas of the medical industry as valuable information could be afforded through the fusion of images that enable the clinical decisions to remain effective. With the aim to render an effective image fusion, this paper concentrates on the Bayesian fusion approach, which is tuned optimally using the proposed Fractional Bird Swarm Optimization (Fractional-BSA). The medical image fusion is progressed using the MRI brain image taken from the BRATS database, and the source images of multimodalities are fused effectively to present an information-rich fused image. The source images are subjected to the Haar wavelet transform, and the Bayesian fusion is performed using the Bayesian parameter, which is determined optimally using the proposed Fractional-BSA optimization. The proposed optimization is the integration of the fractional theory in the standard Bird Swarm Optimization (BSA), which improves the effectiveness of image fusion. Unlike any other existing methods, the proposed Fractional-BSA-based Bayesian Fusion approach renders a good quality and complex-free fusion experience. The analysis reveals that the method outperformed the existing methods with maximal mutual information, maximal peak signal-to-noise ratio (PSNR), minimal root mean square error (RMSE) of 1.5665, 44.0857 dB, and 5.4840, respectively. Graphical abstract Schematic diagram of medical image fusion Medical IF is the significant research domain, which affords the fused image in such a way that this image carries a greater availability of the information content regarding any scene than the information carried by the single source image. Moreover, the concept of fusing multimodality enhances the contents in the image, which increases the reliability and overall information of the image. Thus, the efficient representation of the input data is made through the medical IF such that the physicians are assisted with a wide range of data for effective decision-making. Thus, the paper deals with the medical IF based on the Bayesian fusion approach for which the variable modalities of the image are used. The input image considered is the MRI brain image with four modalities, Flair, T2, T1, and T1C. Among the four modalities, any of the two modalities are considered the source images for fusion. The first step in IF is the generation of the wavelet coefficients, low-low (LL), high-low (HL), low-high (LH), and high-high (HH) using the Haar wavelet transform. Upon deriving the wavelet coefficients, the wavelets are fused based on the Bayesian fusion, which is progressed based on the proposed Fractional-BSA. Once the fused bands are formed, the inverse Haar wavelet transform generates the fused image, and it is significant to note that the IF is performed at the pixel level in such a way that the image quality is assured with a high level of the information for clinical applications. The advantages of the pixel-level fusion are regarding the original measured quantities, which involve directly in the fusion process.