Robust Optimized Structural Feature-Based Transformation Parameter Estimation for Image Registration

Automatic and proficient image registration is a very energizing task. In this article, we propose optimized structural feature-based robust prediction models to predict the transformation parameters toward image registration. Here, scale invariant feature transform (SIFT) is utilized as a feature extraction algorithm, and equilibrium optimization (EO) is utilized to optimize the number of features. Down-sized feature vectors are used as input datasets of the backpropagation neural network (BPNN) and random forest (RF) to fabricate the prediction model. The present investigation exhibits that the proposed technique can robustly estimate different transformational parameters. The comparative analysis of the proposed technique with other methods is depicted in experimental results.