Techniques of developing panorama for low light images

Panorama is concerned with stitching multiple photographic images of the same scene to produce high-resolution image. There are many challenges in case of developing astrophotography with combining images from land and sky dealing with image noise and subject motion by including spatially variant registration steps into the panorama workflow, this will combine several shorter exposures into a lower noise final image without motion artifacts. The method would overcome two major obstacles to generate night sky panoramas: low SNR and motion blur. First the images are segmented into land and sky. Next the locations of probable stars from the star image are extracted. Here extraction of features from night image is challenging task. SIFT algorithm is scale invariant as well as rotation and this is effective in presence of noise than other techniques. Now for astrophotography panorama it requires more features to be extracted. From the comparison of feature extraction techniques, SIFT extracts more features than others and also it is invariant to rotation, illumination, and affine transformation changes, and shows good performance in this case. This star features are then matched between two images which include same feature points. Thus two short exposures get combined. Proper blending technique is applied to remove seam between two combined images. Motion of the stars is compensated by warping the images smoothly using the local transformations and these combined exposures were stitched into a panorama using a spherical projection.