Panorama Generation using Feature based Mosaicing and Modified Graph-cut Blending

Abstract

Panoramic mosaicing has several important area of applications including computer vision/graphics, virtual reality, and surveillance. Different factors like sensor noise, camera motion, and illumination difference affect the quality of the mosaic due to creation of artificial edges (artifacts) between the images in the resulting panorama. This work presents a technique to overcome such problems for the generation of seamless panoramic images. The proposed technique uses Scale Invariant Feature Transform (SIFT) features for image alignment and modified graph-cut (MGC) for blending the seam in the overlapping region between images. The proposed method is tested on various sets of images to show its effectiveness. Comparison with different mosaicing as well as blending techniques show that the proposed method achieves improved results for panorama creation.