Starry Night Panorama with Advanced Feature Extraction and Star Stitching

Abstract

Panoramic photography involves merging multiple photos of the same scene, each with overlapping views, to create a detailed image. When combining astrophotography with panoramic landscapes, challenges arise from image noise and subject motion. To address this, incorporating spatially variant registration steps in the panorama process can merge several shorter exposures into a final image with reduced noise and without motion artifacts. This method tackles two main issues in creating night sky panoramas: low signal-to-noise ratio (SNR) and motion blur.Initially, the images are divided into land and sky segments. Then, potential star locations are identified from a star image. Extracting features from night images is complex, and the Scale-Invariant Feature Transform (SIFT) algorithm is chosen for its robustness to rotation, scale changes, and noise. In astrophotography panoramas, more features need extraction, and SIFT performs well compared to other methods.Next, matching star features between images with common points allows combining two short exposures. A seamless blending technique removes visible seams between merged images. Compensating for star motion involves warping images using local transformations for smooth alignment. Finally, the combined exposures are stitched into a panorama using a spherical projection method.