In addition, there also exist photometric inconsistencies to different extents in overlap regions between adjacent images due to illumination variations and/or different exposure settings. Therefore, those images cannot be precisely aligned in geometry namely, there exist the geometric deviations for corresponding pixels from different images to different extents. Generally, the panoramic camera is comprised of multiple wide-angle or fish-eye cameras whose projection centers are slightly different. However, as we know, most of the street-view panoramic images are captured by a panoramic camera mounted on a mobile platform. In an ideally static scene in which both the geometric misalignments and the photometric inconsistencies do not exist or are not obviously visible in overlap regions, the stitched or mosaicked image looks perfect only when the geometric distance criterion is used. Image stitching is the key technology to produce high-quality panoramic images, which is also an important and classical problem in the fields of photogrammetry, remote sensing and computer vision, which is widely used to merge multiple aligned images into a single wide-angle composite image as seamlessly as possible.
Nowadays, with the development of street-view panoramas, which provide 360 ∘ panoramic views along streets in the real world, the demand for high-quality panoramic images gradually becomes greater. Experimental results on a large set of challenging street-view panoramic images captured form the real world illustrate that the proposed system is capable of creating high-quality panoramas. At last, the Laplacian pyramid blending algorithm is applied to further eliminate the stitching artifacts along the optimal seam lines. After that, the optimal seam lines between adjacent input images are detected via the graph cut energy minimization framework. Then, to lessen the influence of photometric inconsistencies caused by the illumination variations and different exposure settings, we propose an efficient color correction algorithm via matching extreme points of histograms to greatly decrease color differences between warped images.
Therefore, an efficient image warping method based on the dense optical flow field is proposed to greatly suppress the influence of large geometric misalignment at first. Since the input images are captured without a precisely common projection center from the scenes with the depth differences with respect to the cameras to different extents, such images cannot be precisely aligned in geometry. Our proposed framework is comprised of four major steps: image warping, color correction, optimal seam line detection and image blending. In this paper, we propose a unified framework to generate a pleasant and high-quality street-view panorama by stitching multiple panoramic images captured from the cameras mounted on the mobile platform.
0 kommentar(er)
