Fourier-Mellin Transform-Based Registration System Implemented in MATLAB

Resource Overview

A MATLAB-based image registration system utilizing Fourier-Mellin transform algorithm, where executing fm_gui.m launches the demonstration interface. The left-side image panels allow selection of two input images for registration with support for zoom and rotation transformations. The right-side control panel includes "Window" dropdown for filter selection, "Rotation"/"Scale"/"Log-polar" options for interpolation methods, and "REGISTER" button with slider for peak point selection. Registration results display in bottom panels with transformation parameters shown in "Translation"/"Rotation"/"Scale" fields.

Detailed Documentation

This MATLAB-implemented registration system based on Fourier-Mellin transform launches a demonstration interface upon running fm_gui.m. The system implements phase correlation methodology through FFT operations where the left-side image frames allow selection of two input images for registration, supporting scale and rotation transformations through geometric transformation functions. The right-side control panel provides multiple filter windows under the "Window" dropdown menu, while the "Rotation", "Scale", and "Log-polar" options offer different interpolation methods (such as bilinear or bicubic) for respective transformation operations using MATLAB's imrotate and imresize functions. The "REGISTER" button incorporates peak detection algorithm where users can adjust the total number of candidate peak points via slider control before execution. Upon clicking "REGISTER", the system performs Fourier-Mellin transformation with log-polar coordinate conversion and displays the aligned images in corresponding bottom panels. The transformation parameters panel at the bottom quantitatively shows offset, rotation angle, and scaling factors between registered images using text boxes labeled "Translation", "Rotation", and "Scale". This registration system provides an intuitive GUI interface with comprehensive options implemented through MATLAB's image processing toolbox, enhancing registration flexibility and accuracy through configurable algorithmic parameters.