A Novel Image Alignment Technique Leveraging Teaching Learning-Based Optimization for Medical Images

In image registration, traditional optimization techniques are incapable of detecting the optimum value of geometric transformation parameters. To resolve this issue, a novel scheme of monomodal (isomodal) biomedical image registration employing teaching learning-based optimization (TLBO) is proposed. In pre-processing, reference image undergoes gaussian filtering to eliminate noise followed through normalization. During de-noising, contrast between anatomical features of an image is degraded. In order to create the floating image, rigid transformation is employed. These images are aligned by detecting optimum value of rigid transformation parameters (RTP) using TLBO with mutual information (MI) maximization as an objective function. MI and structural similarity index measure (SSIM) are used to evaluate visual quality of registered image. The proposed scheme is tested on several isomodal medical images such as magnetic resonance imaging (MRI) and computed tomography (CT). The value of MI, SSIM increases by 8% and the value of RMSE is significantly reduced from 0.4953 to 0.1306 [1] and 3.7858 to 0.1809 [2] which clearly reveals that proposed scheme is robust and effective as compared with the state-of-the-art methods.