Seasons time-lapse - alignment

This article describes the technical process of aligning hundreds of photos taken from slightly different viewpoints over several years to create a time-lapse. It explains the challenges of varying camera positions, angles, and focal lengths due to different phones. The solution uses OpenCV's ORB algorithm for feature matching and RANSAC for geometric transformation. The author details issues with homography causing distortion and the fix using a simpler 2x3 matrix for translation and uniform scale, plus discarding rotation to prevent horizon drift. This is a practical guide for developers working on image alignment and computer vision projects.