Research on intelligent registration methods for multidimensional visual communication projection transform images

This article presents an intelligent image registration method tailored for images involved in multidimensional projection transformations in visual communication, which often suffer from blurred contours and incomplete shape alignments post-registration. The proposed approach comprises four integrated stages: (1) wavelet transform-based denoising for feature preservation, (2) edge detection using the Laplacian of Gaussian (LoG) operator, (3) contour matching via Fourier descriptors, and (4) projection and affine transformation for precise alignment. The method was evaluated on two high-resolution landscape datasets (Elephant Trunk Hill and Potala Palace, 1,920 × 1,080), achieving 99.1% accuracy and a root mean square error (RMSE) of 0.31. Compared to four state-of-the-art methods (DOI 10.1007/s10586-023-03974-3 ; DOI 10.1109/jsen.2023.3314608 ; DOI 10.1109/ACCESS.2023.3264968 ; DOI 10.1111/phor.12402 ), our approach demonstrated superior accuracy and execution speed. The technique is feasible for near real-time application on standard hardware (Intel i7 CPU, 16 GB RAM) and offers robustness without requiring supervised learning. This work contributes a clear, interpretable, and efficient framework for image registration.