High-Precision Image Registration and Fusion Based on Multi-Scale Features

Image registration is a core technology in the field of image processing, widely applied in areas such as medical imaging analysis and remote sensing monitoring. This paper proposes a high-precision image registration method based on multi-scale feature extraction. Firstly, a multi-scale image pyramid technique is employed to integrate local and global features, effectively enhancing the accuracy of image registration. Building on this, an optimized strategy combining rigid and non- rigid registration is introduced to address challenges such as rotation, scaling, illumination variations, and noise in images, further refining the registration process and ensuring high- precision alignment in complex scenarios. Experimental results demonstrate that the proposed method achieves more accurate alignment in both high-resolution and low-quality images, showcasing strong robustness and broad application potential.