A study of image registration and color blending for close-range and aerial images

In this paper, an image mosaicing system is introduced. The proposed system consists of three main steps, feature matching, image registration, and color blending. In the first step, the approach of scale-invariant feature transform is adopted to describe the extracted features for matching. In the step of image registration, a least-square-based approach is adopted to align the matched features obtained in the first step. In the last step, the images are seamlessly mosaicked together by a multi-band color blending algorithm. The experimental results on close-range images and large-scale aerial images show that the proposed system can smoothly blend the colors on boundaries and thus seamlessly mosaic the images.