Ghosting Elimination with A* Seam Optimization in Image Stitching

This paper presents an algorithm for eliminating the blurred image regions during image stitching. Since the camera is close to the objects when capturing the images in the indoor environments, stitching the images in the indoor scenes easily generates the blurred ghosting region. Here the reasons of blurred region generations can be categorized into two kinds. First, the slight camera vibration may result in the whole images unclear over a short span of frames. These unclear frames can be filtered out by distinguishing the clearness of the edge images between different scales. The other reason for originating the blurred region is that several image frames contain the same scene from different views. There may exist the inconsistent structures and observations in these different views. The color blending is performed to let the seam between aligned images smooth, however, it also makes these inconsistent region blurred, which is called the ghosting effect. Here, an optimal seam selection algorithm is proposed by utilizing the A* search algorithm to find a segment path away from the inconsistent overlap region and then remove the ghosting effect. The experimental results validate that the proposed algorithm can obtain the mosaic image with good quality and clear resolution.